University of Colorado, Boulder
CU Scholar
Undergraduate Honors Theses
Honors Program
Spring 2016
The Inefficacious Electorate: Political Powerlessness
and Vote Choice
Ryan W. Bell
Student, [email protected]
Follow this and additional works at: http://scholar.colorado.edu/honr_theses
Part of the Comparative Politics Commons
Recommended Citation
Bell, Ryan W., "The Inefficacious Electorate: Political Powerlessness and Vote Choice" (2016). Undergraduate Honors Theses. Paper
1213.
This Thesis is brought to you for free and open access by Honors Program at CU Scholar. It has been accepted for inclusion in Undergraduate Honors
Theses by an authorized administrator of CU Scholar. For more information, please contact [email protected].
The Inefficacious Electorate
Political Powerlessness and Vote Choice
Ryan W. Bell
Political Science Departmental Honors Thesis
University of Colorado, Boulder
Defended March 18, 2016
Defense Committee
Dr. Fitzgerald, Thesis Advisor
Political Science
Dr. Adler
Political Science
Dr. McKinnish
Economics
Abstract
Non–traditional parties have had newfound success across European democracies contributing to increasing electoral volatility. Drawing from the literature on social movements,
voting for an unconventional party can be a form of contentious politics. Support for non–
mainstream parties is the institutional manifestation of the factors that contribute to social
movements. Perceptions of political power and grievances extend beyond non–institutional
actions, shaping how individuals cast their ballot, with the inefficacious and dissatisfied voters engaging in electoral disorder. An attitude–ideology approach to non–traditional party
support is tested with Bayesian inference using Dutch Parliamentary Election Studies from
1977 to 2006. Voters’ perceived influence in and satisfaction with government are important
factors for deciding between a mainstream and outsider party, supporting a connection between unconventional behavior outside institutions and within the electoral process. When
at the ballot box, non–ideological factors have the potential to alter party choice, with the
alienated and disaffected portion of the electorate behaving differently than the efficacious
and satisfied.
1
Contents
1 Introduction
5
2 Motivating Voters: Participation and Choice
7
2.1
Ideology and the Vote . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2
Non–Ideological Components of Voting . . . . . . . . . . . . . . . . . . . . . . . . 14
3 Political Efficacy and Expressive Motivations
17
3.1
Power Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2
Social Movements and Party Choice . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3
Vote Switching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.4
Mainstream versus Outsider Support . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.5
The Role of Ideology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4 Data and Methods
31
4.1
The Netherlands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.2
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2.1
Dependent Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2.2
Independent Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2.3
Variables Across Elections . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.3
Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.4
Model Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5 Results
5.1
48
Voters’ and Parties’ Ideology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.1.1
Vote Switching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.1.2
Mainstream Voting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.1.3
Party Choice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6 Discussion
6.1
57
Vote Switching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2
6.2
6.1.1
Vote Switching and Individual Influence . . . . . . . . . . . . . . . . . . . 59
6.1.2
Vote Switching and Satisfaction . . . . . . . . . . . . . . . . . . . . . . . . 61
Mainstream Voting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.2.1
Efficacy and Voter–Party Power Dynamics . . . . . . . . . . . . . . . . . . 65
6.2.2
Satisfaction and Affective Motivations . . . . . . . . . . . . . . . . . . . . 66
6.2.3
Left and Right–Leaning Voters’ Relation with their Parties . . . . . . . . 67
6.3
Party Choice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
6.4
Electoral Earthquakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.5
A Protest Voter? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
7 Conclusion
7.1
73
Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
8 Appendix
78
8.1
Appendix A: Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
8.2
Appendix B: Vote Switching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
8.3
Appendix C: Mainstream Voting . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
8.4
Appendix D: Frequentist Regression Results . . . . . . . . . . . . . . . . . . . . . 92
8.5
Appendix E: MCMC Diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
8.6
Appendix F: Diagnostic Test Statistics . . . . . . . . . . . . . . . . . . . . . . . . 99
3
List of Figures
1
Efficacy and Voting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6
2
Attitudes and Behavior Across Time . . . . . . . . . . . . . . . . . . . . . . . . . 37
3
Parties’ and their Voters’ Left–Right Ideologies . . . . . . . . . . . . . . . . . 49
4
Vote Switching Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5
Expected Probability of Vote Switching . . . . . . . . . . . . . . . . . . . . . . . 51
6
Mainstream Voting Regression Table . . . . . . . . . . . . . . . . . . . . . . . . . 54
7
Expected Probability of Mainstream Vote . . . . . . . . . . . . . . . . . . . . . . 54
8
Posterior Predict Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
9
Posterior Distribution of Regression Coefficients . . . . . . . . . . . . . . . . . . . 89
10
Efficacy and Satisfaction of Left and Right–Leaning Voters . . . . . . . . . . . . 90
List of Tables
1
Posterior Inclusion Probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2
Changing Party Loyalties: Past and Present . . . . . . . . . . . . . . . . . . . . . 50
3
Variables by study Year . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4
Changing Party Loyalties: Intension and Actual . . . . . . . . . . . . . . . . . . . 82
5
Vote Switching Regression Table with Policy Distance . . . . . . . . . . . . . . . 82
6
Hesitated Party Choice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
7
Mainstream Voting Regression Table with Left–Right Relative Extremeness . . . 84
8
Mainstream Voting Among Liberal Parties . . . . . . . . . . . . . . . . . . . . . . 84
9
Mainstream Voting Among Conservative Parties . . . . . . . . . . . . . . . . . . 85
10
Mainstream Voting by Election, 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
11
Mainstream Voting by Election, 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
12
Posterior Predictive Distribution Misclassification . . . . . . . . . . . . . . . . . . 90
13
Frequentist Regression Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4
1
Introduction
Elections in Europe have experienced increasing unpredictability, while outsider parties have
had newfound success. European Parliament elections in 2009 and 2014 saw sizable increases in
Eurosceptic MEPs; National Front in France has had recent success in local elections, setting the
groundwork to contest the presidential election; the Greek electorate, rejecting the established
New Democracy and PASOK, elevated SYRIZA to the majority party; the Spanish parliament is
the most fragmented in history, with the traditional left and right parties losing to the upstart
PODEMOS and Citizens parties; Law and Justice has become the dominant force in Polish
politics; while Alternate for Germany has won representation in German states, and is poised
to make gains in federal elections. Across parliamentary democracies, establishment party
dominance faces increasing challenges from the left, right, and more idiosyncratic movements.
New parties have emerged and gained rapid electoral success, directly challenging the status
quo of mainstream party control.
Why have non–traditional movements become increasingly successful? While anti–
immigrant and populist politics are the most prominent, they are by no means the only. These
new movements across Europe span the ideological spectrum. Far–right politics have been the
focus of an expanding literature, but there is less work on the growth of the far–left and other
party types; and works on the broad challenges to traditional parties is less plentiful. Have
European voters’ political orientations shifted and their ideology changed, or do the origins of
outsider support lie elsewhere? This paper proposes that the decline in mainstream support
may be a non–ideological phenomenon, rooted in political attitudes and the ways in which voters
see themselves in relation to government. External efficacy can drive citizens to the polls, but
once they have arrived, can it influence the party for which one will cast their ballot? Will the
inefficacious behave differently than the politically empowered, and will those dissatisfied with
government act differently than those that are satisfied?
In non–institutional contexts, disaffected and alienated individuals are more likely to
engage in non–traditional, more contentious actions, such as attending demonstrations and
participating in riots, but are there institutional analogs of these behaviors? I apply the work
on social movements to a system–focused context, testing the effects of political attitudes on
electoral behavior. While vote choice and social movements have mostly remained distinct
5
literatures, there may be a greater connection between the two. Voting is more individual,
lacking the collective action component of movements, but the motivations of both phenomena
may share common roots. Treating traditional voting as ideologically–based decision–making,
in which one engages in spatial behavior, I examine non–traditional voting: situations in which
ideology is not the main concern behind an individual’s party choice. Considering support
for outsider parties as a form of unconventional vote choice, then its cause will follow that
of non–traditional, non–institutionally–focused participation, together forming a repertoire of
unconventional behavior.
In the political participation literaFigure 1: Efficacy and Voting
ture, individuals’ expressive desires and attitudes have contributed to the understand-
2.0
Efficaciousness of Voters and Non−Voters
ing of political action, both institutionally and
Voters
Non−Voters
efficacy and satisfaction to an ideological–
understanding of party choice has the poten-
1.0
Density
1.5
non–institutionally focused. The addition of
tial to replicate the successes in the participa-
0.5
tion literature, furthering the understanding
of political decision–making. Removed from
0.0
parties’ relative position with respect to a vot0.0
0.2
0.4
0.6
0.8
ers’ liberal–conservative placement, how will
1.0
Efficacy Score
an individuals’ perceptions of their political
power and their dissatisfaction with government influence their actions? There is a strong role
for an individual’s belief in the effectiveness of their actions. If one does not trust that their
political actions will have consequence, then they are unlikely to take such action.
Electoral disorder does not strictly fit within traditional conceptualization of voting
behavior or social movements. Due in part to the assumption that those who view themselves as
politically powerless will not vote, inefficacious voters’ behavior has seen less analysis. Likewise,
impacts of inefficacy on party choice has not been researched in the social movement context, as
one’s vote is more individualistic in nature, removed from mobilizing and organizational aspects
of a movement. Non–traditional electoral action is a more personal, singular, and anonymous
form of un–conventional behavior. Densities of the efficacy score of those who intend to vote
6
and non–voters can be seen in 1. Voters, as the work on participation predicts, are more
empowered, while non–voters are substantially more alienated. Most voting respondents are
rather efficacious, scoring between .8 and one; however, the third most scoring category is
comparatively lower, ranging from .2 to .4. Although those alienated will vote in lower numbers
than the empowered, there is still a substantial portion of low–scoring Dutch Parliamentary
Election Study (DPES) respondents that still intend to vote. This project focuses on this
inefficacious component of the electorate, that, despite lack of efficacy, are still driven to the
polls. The central question is whether their inefficacy will affect how they cast a ballot.
2
Motivating Voters: Participation and Choice
Work on participation, and voter turnout in particular, has benefited from the inclusion of
individuals’ attitudes. Expanding motivations from strictly economic factors to include variables
such as civic duty has created a more full understanding of the decision to cast a ballot. Anthony
Downs introduced Rational Choice to voting behavior in “An Economic Theory of Political
Action in a Democracy” (1957). Downs theorized voters as utility-maximizing actors that select
candidates based on which will provide the most personal benefit; and candidates for office were
suggested to be motivated by income, power, and prestige goals. In order to win elections, parties
create policies that maximizes a majority of the population’s utility, and therefore gaining the
most votes (Downs 1957). When deciding whether or not to vote, and individual makes their
choice based on the utility function: R = pB − c. In the voter’s calculation, p is the probability
that one’s vote will be pivotal, determining the outcome of the election; B is the benefit from
the preferred candidate winning; and c is the cost of voting. When the probability of the desired
outcome multiplied by the benefit of that outcome is greater than the cost, the individual will
vote. That is, if the benefit gained from voting is positive, R > 0, then the individual will
turnout on election day. While an economic understanding provides a theoretical basis for the
decision to participate, in its original form, the theory produces an inconsistency. The strict
cost–benefit approach is unable to explain the most fundamental political action in democracies:
given Downsian assumptions, it is irrational to vote. The negligible probability of one casting
the deciding vote for their most preferred candidate is overshadowed by the costs; and when
adjusting the utility function to an individual’s perceived probability of casting a decisive vote,
7
the costs still outweigh the benefit (Ferejohn & Fiorina 1974, Nurmi 1999, Ordeshook 1976). In
order the rectify the inconsistency, one must incorporate non–economic factors in the electorates
evaluation of the benefits gained from voting.
One proposed adjustment to Downs’ assumptions is cost minimization. Alternative
to voters maximizing their benefit, individuals may act in a manner that will minimize their
potential loses. A would–be voter is not motivated by the belief that their vote will have
tangible effects; instead, one will seek to reduce potential losses from not participating (Ferejohn
& Fiorina 1974). Perceived harms from the least preferred candidate winning drives a cost–
minimizing individual to the polls in order to reduce the probability that the disliked candidate
wins. While it is improbable that one will cast the deciding vote for their supported party, the
costs of voting are less than not participating in an election in which the most disliked party wins.
The use of Prospect Theory, especially loss–aversion, produces a more accurate understanding
of participation than the original Downsian assumption (Quattrone & Tversky 1988). However,
loss–minimizing assumptions are still unable to explain action lacking a clear economic incentive.
A variety of solutions have been proposed to overcome the paradox of voting. While
many focus on a strictly economic basis, there is a large body of literature that make use
of political attitudes. One such attitudinal approach was that of Riker and Odershook, whose
work modified Downs’ assumptions with the addition of a new, non–economically based variable.
Besides the probability of creating the desired change, a voter may gain utility, and therefore
motivated by “a sense of civic duty.” Riker and Odershook altered the function to include a
citizen’s perceived obligation to vote with the D term (1968): R = pB − c + D. The addition of
the D term captures, as defined by Campbell, Gurin, and Miller, “the feeling that oneself and
others ought to participate in the political process, regardless of whether such political activity
is seen as worth while or efficacious” (1954, 194). Regardless of the cost of casting a ballot,
citizens of a democratic society have an obligation to vote. When the strength of this sense
of duty, combined with the perceived benefits, outweigh the costs, an individual will vote. As
opposed to the expected benefit of political action, the D term is a sort of expected cost of not
participating, i.e. the cost of not upholding one’s duty. Although “a sense of civic duty” is not
an economic consideration and cannot be quantified in a strict cost–benefit context (Ferejohn &
Fiorina 1974), the inclusion of the subjective attitudinal component makes the Rational Choice
framework consistent (Nurmi 1999, Quattrone & Tversky 1988, Whiteley 1995).
8
Individuals’ motivations, their goals, define the factors that influence political participation; in order to understand what drives one to the polls on election day, and what influences
vote choice, one must understand voters’ expectations regarding their participation and support
for a particular party. One such variable is the extent to which an individual is able to influence
the system. If a voter feels their vote will be more effective in bringing about desired change,
in achieving their goals, there is a greater motivation for action (de Moor 2014). Coined by
Campbell, Gurin, and Miller in The Voter Decides, political efficacy is defined as “the feeling
that individual political action does have, or can have, an impact upon the political process,
i.e., that it is worthwhile to perform one’s civic duties” (1954, 187). In a similar perceptive,
Mokken (1971) described political efficacy as the sense that one can influence the system. However, evaluations of the governments willingness or ability to implement citizens’ desires are
independent of individual factors. In additional to external components, such as government
responsiveness, there is a self–focused dimension to efficacy. One’s own ability, or competency
to affect change—separate from the government’s receptiveness to input—defines one’s level of
internal political efficacy (Coleman & Davis 1976). Both components of efficacy have substantive implications for participation, and perception of an individual’s political power can used to
understand individuals’ motivations for such action.
Perceptions of control over government and system responsiveness have substantial
implications for political mobilization. As external efficacy increases, so does the chance of
institutionally–focused participation, such as voting (Campbell, Gurin & Miller 1954, de Moor
2014, Pattie & Johnston 1998, Klandermans 1996, Zimmerman, Israel, Schulz & Checkoway
1992, Whiteley 1995). When the government is seen as more receptive to its citizens desires, then
there is a greater perceived benefit in giving one’s input. If the government is seen as unwilling
to address an individual’s concerns, then they will use more confrontational means outside
of established institutions. For example, efficacy is a significant factor in protest attendance
or rioting participation (Craig 1980, Craig & Maggiotto 1981, Gamson 1968, Nachmias 1974,
Paige 1971, Pollock III 1983, Seligson 1980, Wolfsfeld 1986). As the government does not
address concerns when institutional routes are used, one must take extra–system actions to
redress grievances. A perception of government obstinance will lead one to use protest as a
mean to address their concerns (Klandermans 1996). The perceived power relationship between
an individual and their government is consequential for political behavior, whether channeled
9
through existing institutional corridors or extra–system means. In deciding whether one will
vote, or how they will participation, an individual considers the effectiveness of their chosen
action on exerting influence within government.
Efficaciousness motivates the individual, in that the perception that one’s participation
is likely to create the desired change drives action. In the context of voters’ utility functions,
efficacy can offer an alternative interpretation to the p term. Instead of the probability of
determining the outcome, internal and external efficacy can be seen as the extent to which an
individual senses they can bring their political desires to fruition. Replacing an objective or
subjective measure of the probability of casting a pivotal vote, efficacy accounts for general
influence over government. Beyond the ballot box, an individual must believe that their most
preferred party will uphold their electoral promises after winning an election. One may decide
an election with their vote, but if the winning party does not represent the voter thereafter,
the benefit received would be equal to that of not participating. Perhaps equally important
to affecting an election outcomes is an individuals influence within the most preferred party.
The efficacious sense that their participation will be more likely to create the desired outcome,
increasing the probability of receiving the benefits from their action, and therefore motivating
such actions (Finkel & Muller 1998). Likewise, inefficacy ferments the belief that one’s action
will not have a meaningful impact for bringing out change, which reduces perceived benefits
of political action and decreases the likelihood of voting. Due to its influence on the expected
probability of receiving benefit, inefficacy can perpetuate inaction.
Use of economic or attitudinal components emphasize different motivations: instramental and expressive, respectively.
Strict economic voting models focus on instrumental
considerations—voting meant to determine the winner of elections, with the calculus based
in probability of influencing the outcome (Achen 2006, Franklin, Niemi & Whitten 1994). In
a voter’s utility function, the p term represents one’s belief that their vote will be pivotal. Alternatively, although not mutually exclusively, attitudinal formulations of Downs’ assumption
make use of voters’ expressive desires. Motivational assumptions are relaxed from a strictly
economic paradigm (Harsanyi 1969), including elements such as upholding one’s civic duty.
The use of efficacy is not oppositional to Rational Choice; instead, political efficacy can be
seen as a replacement of the p term with a subjected measure of the expected benefits from
participation(Finkel & Muller 1998). In a comparison of efficacy and economic–based motiva-
10
tions for voter turnout, both items performed equally well (Pattie & Johnston 1998), reinforcing
the importance and compatibility of non–economic and economic elements behind voters’ decisions. In the 1987 British elections, Franklin, Niemi, and Whitten (1994) found that while some
voters behaved strategically, selecting the party best able to win, others chose to vote expressively, without concern for possible pivotal consequences of their choice. They selected a party
to signal or warn their most preferred party, or voted in a certain way to register support for another party’s policies in hopes that the voters’ most preferred party would adopt those positions.
Voters do not cast a ballot believing that they will determine the election; instead, they are
motivated by showing support, or turnout on election day in accordance with social norms, upholding their civic duty. Assuming that voters follow Bayesian updating—combining new with
old information and adjusting beliefs—and act expressively improves model performance over
strict instrumental approaches (Achen 2006). Overtime, strict economic analysis of political
action has been relaxed to include attitudinal motivation effects and expressive considerations
into the understanding of participation. While still making use of Downsian utility–maximizing
voters, the utility function has been expanded to include attitudinal variables, separate from
determining electoral outcomes and tangible benefits.
2.1
Ideology and the Vote
Concerned directly with vote choice, a spatial understanding of party selection developed from
Downs economic approach. In the two–party case, the ideology space is unimodal, centered
about the median voter; in multi–party, parliamentary democracy, the ideology space is multi–
modal, in which there is a party to fill the space for each mode (Downs 1957). In the unimodal
case, both parties converge towards the median voter in order to capture a majority of the
electorate, while in multi–party systems, the candidates are distinctly separated, remaining
within their neighborhood of the ideology space. Where parties are effectively bound to an
ideological area, they attempt to maximize their vote share of like–minded voters. As absolute
majorities may be impossible, parties seek a plurality of the electorate and the ability to form
coalitions. When deciding which organization to support, voters can engage in hard issue
voting, in which decisions are reached through a more costly conscious calculation, or one may
use soft issue voting, based in heuristic gut responses that are policy ends–oriented and more
symbolic (Carmines & Stimson 1980). Hard or soft issue voting determines the amount of time
11
and energy dedicated to party selection. Informed from an economic approach, spatial voting
assumes utility maximization behaviors of voters. Ideology and or policy positioning is the
determinant of vote choice, in which one votes correctly if they support a party most closely
aligned to themselves (Jessee 2012). As with the strictly–economic analysis of voter turnout,
a strictly–ideological approach to vote choice may present an incomplete understanding; while
maximizing ideological proximity provides a theoretical basis, the exclusion of attitudinal factors
reduces the ability to explain voting behaviors.
Formal spatial models use Downsian assumptions to form a deductive framework of
voters’ party selection. The models assume that each individual votes sincerely: an individual’s
goal is to vote correctly (Ordeshook 1976). While spatial models offer a simplifying framework
that still provides a large amount of explanatory power, the parsimony runs the risk of neglecting important factors in vote choice. Individuals tend to support the party most closely
aligned to their preferences, but spatial behaviors cannot explain non–ideologically–based actions. Likewise, policy distances cannot determine which organization an individual will choose
when they are in the two parties’ ideological midpoint. In American presidential elections, an
ideology–only model’s predictive ability for those equally close to the two candidates is equivalent to that of a coin flip (Jessee 2012). For a deeper understanding of individual behavior,
spatial models have the potential to replicate the success of developments to research on voter
turnout.
Similar to work on participation, formal spatial models have expanded the utility–
maximization criteria, incorporating expressive motivations to examine the theoretical consequences. In party choice, there is a tension between signaling preferences and instrumental
considerations: when one’s most–preferred party is not the most politically powerful, which do
they choose? One may cast a strategic vote for the party with the greatest chance of winning, or
choose a party that is most closely–aligned to the voter’s ideology. If voting correctly, one would
select the party more aligned with themselves, but pivotal concerns would motivate a vote for
the more powerful organization. While behaving in an attempt to influence future ideological
positioning, a signaling voter puts aside their spatial concerns in the current election in order to
alter parties in future elections (Shotts 2006). Likewise, one may engage in “signal jamming”
behavior, which is not voting for the most preferred candidate in order to prevent electoral
landslides. Jamming functions as a strategy for balancing victory for the desired candidate
12
while preventing an over–sized electoral mandate, which can lead to candidates overstepping
(Myatt 2015). Individuals base their behavior on the expected action of others, attempting to
gauge and regulate the probable margin of victory. If a voter assumes that a party will adjust
their position after suffering electoral losses, a voter has a clear path to send a message to the
party. When dissatisfied, one may engage in protest: they will select a party other than the
most preferred in order to communicate their disapproval (Kselman & Niou 2011). When the
electorate has lost faith in the candidates, they will choose the “lesser of two evils” and punish
the most relativly worse candidate (Levin & Eden 1962). Not necessarily motivated by ideology
or enthusiasm for one party over the other, the voter bases their decision in a cost–minimizing–
like fashion—one is motivated to prevent the most disliked candidate from winning. While
formal modeling has explored theoretical implications of non–strictly–ideological behaviors, the
transition has been slow to applied models.
Empirical spatial models have been slower to adopt attitudinal elements than their
formal counterparts. Following the assumptions of correct voting, survey analyses focuses on
the difference between parties’ positioning and voters’ stance in the ideology space, controlling for contextual variables (Fennema 1997, Quinn & Martin 2002, Van der Brug, Fennema
& Tillie 2000). Work on spatial models uses a Downsian political market place, in which candidates supply policies to meet citizens ideologies, to analyze success, or failure, of new and
anti–establishment movements across Europe. In the case of electoral volatility, voters will
change their party loyalty if either they, or their party, change their ideological stance. As a
consequence, outsider movements are explained with changing ideology and preferences of voters. The rise of outsider political movements is driven by a demand for their policies, currently
un–supplied by existing parties. Analyses have found that contextual factors—supply–side elements affecting ability to provide an outsider ideology—explain a fair amount of variation between countries. Across democracies, the shape of political institutions and electoral regulations
influence success of non–established parties, determining the entry costs and voters’ strategic
incentives. However, individual–level, demand components has seen less analysis (Van der Brug
& Fennema 2007, Golder 2016). On the demand side, there is a consensus on the importance
of an existing anti–immigrant sentiment for the success of extreme–right parties (Arzheimer &
Carter 2006, Arzheimer 2009, Coffé 2005, Fennema 1997, Thijssen & Dierickx 2001, Van der
Brug et al. 2000, Van der Brug & Fennema 2003, Van der Brug & Fennema 2007). Individuals
13
that desire more conservative immigration policies select a radical–right party in order to bring
about the desired change.
Outside a strict supply and demand setting, the limited work on socio–demographic
and attitudinal components is mixed. One such case of conflicting results concerns the different
success of the Flemish Vlaams Blok and the absence of such a Wallonian movement (Coffé 2005,
Thijssen & Dierickx 2001). While attitudinal factors are an influence in Flanders, their effects
are less prominent in Wallonia. Likewise, class interest and group membership is less clear
on outsider support. For vote choice in general, once–powerful socio–demographic predictors
are increasingly less potent. As old cleavages fade, as is the case of de–pillarization in the
Netherlands, traditional variables behind party support are losing relevance (Van Holsteyn &
Irwin 2003, Inglehart 1981). Given an individual’s demographics, such as age, religion, and class,
one is now less able to predict vote choice. An economic–based understanding of vote choice is
less able to explain post–materialist movements. There is a large body of work on radical–right
success, but work on the root cause of these types of movements broadly is less plentiful; when
and why they appear, the variability across countries and variability across parties, as well as
the emergence of other movements such as the radical left, populist or anti-establishment, and
Euroscepticism is uncertain (Erlingsson & Persson 2011, Hix & Marsh 2007, Spoon & Hobolt
2010, Ignazi 1996, Van Holsteyn & Irwin 2003, Zaslove 2008, Pahre & Radziszewski 2006). It
remains unclear what underpins the broad success of anti–establishment politics, and whether
there are common roots across the heterogeneous new movements.
2.2
Non–Ideological Components of Voting
Adoption of an expressive voter approach, explored in formal models, has the potential to introduce a more full understanding of voting. As Downsian participation models have benefited
from attitudinal elements, spatial models may equally benefit. There are factors influencing
political decision–making exogenous of ideology; behaviors of an alienated electorate cannot be
explained solely through a spatial approach. Dominance of establishment parties in European
elections has been decreasing: levels of partisan identification have been falling, and dissatisfaction and skepticism towards traditional politics has been rising (Dalton 2000, Ignazi 1996). New
political movements across the spectrum have gained prominence in parliamentary democracies
14
throughout Europe. Neoliberal movements were the first to emerge in post–industrial Europe,
formed as a reaction to voters’ inefficacy with respect to traditional means of participation and
their lack of influence within party structures; these “New Politics” parties served as a new
form of expression (Ignazi 1996). Following the New Politics, “post–industrial extreme–right”
parties grew, seeking to de-legitimize the system. Both movements have origins in the conflict over changing values as well as an individual’s relationship with government (Ignazi 1996).
With parties moving towards the median voter and ideological center to capture the electorate,
new parties form on the right and left periphery, in a Downsian fashion, filling neglected ideological space (Bale 2003). Outsider political parties were given openings on the ideological
spectrum, but there is more underlying the modern political shifts than ideology. Broadly,
anti movements have become increasingly successful. Relational outsider movements—those
beyond the traditional mainstream ideological space—work to undermine and polarize electoral
politics (Capoccia 2002). Populism grew from post–industrialism and the perception of a disconnect between the people and politics. The govern–governed break gives democratic deficits
more salience, forming opportunity structures for populist movements and the success of us–vs–
them appeals (Barr 2009, Zaslove 2008). Alienated from the establishment, resentment towards
the status quo motivates voters to support non–traditional politics. And growing perception
of democratic deficits leads to growing electoral instability (Keman 2014). In a similar vein,
one’s policy preferences might remain stable, but the perception of how parties govern changes.
One may sense that the mainstream has engaged in de–politicization, colluding to suppress
contentious issues and reducing competition between the established organizations; the party
works less for its members and the country, instead ensuring its own survival in an Iron Law
of Oligarchy–like fashion (Katz & Mair 1995, Katz & Mair 2009, Michels 1915). While the
individual’s ideology has remained constant, a perception of the cartelization of politics drives
the individual to unconventional political behaviors. Where party collusion prevents government from addressing issues of interest, the individual will change their party loyalty, instead
supporting an anti party.
Accompanying the emergence of new policy questions are changes to values orientations. In the modern era, economic scarcity has become less prevalent, shifting voters’ focus to
non–material concerns. As one does not have to worry of first–order needs, they progress to
higher–orders, such as self–expression. The ideology space can be dynamic, shifting electoral
15
politics and voter loyalties through shifting orientations, but there are non–spatial factors that
should not be neglected. Post-material values have been increasing throughout Europe. Old
cleavages are decreasing while new post-material conflict over government involvement in the
economy, welfare, and social policy grow (Inglehart 1981, Inglehart & Flanagan 1987). While
left–right self placement is a predictor for party choice (Middendorp 1992), a liberal–conservative
scale alone is insufficient for explaining voting behavior. In the Netherlands, Lijst Pim Fortuyn
(LPF) illustrates the failures of old models based on social cleavages and ideology to explain
electoral choice (Van Holsteyn & Irwin 2003). Support for LPF was not restricted to certain
demographics, instead gaining support across the Netherlands and drawing from all ideologies.
Previous work has found that anti–immigrant parties are evaluated using the same criteria as
traditional organizations (Fennema 1997, Van der Brug & Fennema 2007). While spatial behaviors may align between the outside and mainstream parties, ideological factors by themselves
do not incorporate post–materialist considerations, such as expression and democratic ideals of
citizen control and representation.
Work merging an expressive and instrumental understanding—although more limited—
has been fruitful. A combination of spatial and social characteristics of voters in the Netherlands
performs better than models with ideological variables alone (Quinn, Martin & Whitford 1999).
Focusing citizens’ on evaluations of government, previous research has examined the implications
of incumbent affect: an individual’s perceptions of the current governing officials, independent
from evaluations of the political institutions (Muller & Jukam 1977). In Spanish national elections, the position of parties in the issue space is not an adequate explanation of vote choice.
An individual is not only concerned with their ideological proximity to a party, but also with
the perception of a party’s ability to effectively govern and a basic level of trust (Labzina &
Schofield 2015). Comparing voters’ support for extreme–right versus conventional parties, there
is clear attitudinal distinction, in which the extreme–right voter exhibits greater dissatisfaction
(Arzheimer & Carter 2006). In Belgian elections, affect and trust improve the understanding
of the differences in Flemish and Wallonian parties. Distrustful voters, when given the opportunity, defect from traditional parties and engage in elite–challenging behaviors (Hooghe,
Marien & Pauwels 2011). In an environment with emerging parties and anti–establishment
sentiments, models combining ideology and attitudes provide more in–depth understanding of
individuals’ political decision–making. Voters identify with party families that share a similar
16
ideology; however, ideological proximity alone cannot explain the support for an outsider versus
mainstream party member.
The current framework of understanding vote choice is an instrumental, ideological
approach. Changing dynamics in multi–party elections can be contributed to shifting ideologies; however, the emergence and success of varied and ideologically unique movements and the
broad support of outsider parties is not fully understood within this context. Preexisting demand for new movements’ ideology and spatial concerns contribute to outsider vote choice, but
provide too parsimonious an operationalization. The approach to individual electoral choices
requires more breadth. The utility–maximization framework of political participation began
restricted by narrowly–defined motivations, creating a paradox; incorporation of “a sense of
civic duty,” and further inclusion of attitudinal components created a consistent explanatory
framework. Movement towards an understanding that emphasizes individuals’ perceptions and
their expressive motivations of voters gives a more broad understanding of participatory action. Adding attitudinal elements to voters’ evaluation of parties has the potential to further
the understanding of party selection. While dissatisfaction with government has been used in
explanations of outsider political support (Arzheimer 2009, Erlingsson & Persson 2011, Muller
& Jukam 1977), the implications of the power relationship between an individual and the government have been less researched. Political efficacy, the sense of individual influence over
government, is a foundational attitude in democratic societies. Efficacy has been broadly applied to participation research, but its implications for how one selects a party remains relativly
untested. This paper incorporates expressive considerations into vote choice, examining the
effects of perceived democratic deficits and government unresponsiveness on party selection.
3
Political Efficacy and Expressive Motivations
Voting is an expressive act, but what exactly are voters expressing? Through elections, citizens
can voice their ideological preferences or support for advantageous policies. More fundamentally, the vote can be an expression of a citizen’s disaffection or an avenue to address political
powerlessness. Ideals of system responsiveness form the foundation of democratic norms. And,
even in an autocratic system, symbolic gestures of citizen control are important for regime
success (Coleman & Davis 1976). Individuals are not inspired to action and expend resources
17
on participation because their ideology compels them to action; involvement originates in the
notion that action can bring the desired outcomes to fruition and that the expression will be
successful. The belief that one’s action can have a meaningful impact is a necessary condition for
such action. As a result, one will consider their influence over the system before participation.
If one’s control over government is less than the desired level, an individual will first focus on
restoring their level of influence. Perceptions that action will result in the desired government
response is the foundational motivation for political action. A basic level of efficaciousness is
a prerequisite for higher–order expressions of policy preferences and instrumentally–motivated
decisions. Beliefs that one’s vote will be cast without impact, or that the government will not
respond, can discourage participation and move one to act to in a restorative fashion.
A favorable outcome of one’s political action manifests in attitudes of one’s own ability to create change, improving sense of government’s receptiveness. This perception mediates
emotional responses. When faced with policy threats, those confident in their ability to influence government are motivated to suppress the threat, encouraging political participation
(Valentino, Gregorowicz & Groenendyk 2009). A reciprocity exists between efficacy and action.
In an electoral context, participation, such as voting or campaigning, increases efficacy; the increased efficacy in turn raises the probability of participation, contributing to habitual patterns
(Finkel 1985, Valentino et al. 2009). When political action is successful—such as supporting the
winning candidate—efficacy increases more than in the case of an unsuccessful action (Clarke
& Acock 1989). Engaging with the political system has the potential to produce a sense that
the government is more responsive, as well as increase an individual’s perceived competency
regarding political action. However, if an individual senses that the government has become
unresponsive, the sense of inefficacy, or powerlessness alienation, will create two motivational
sources for action: efficacy and affective responses.
3.1
Power Dynamics
Within democratic institutions, there is a balance of power between individuals’ influence and
government’s mandate. Representatives are bound through elections and democratic norms
to the voters, but also possess their own agency to pursue policy that may not align with
electors preferences. The relationship between a voter and their party forms a principle–agent
18
problem. Tension between citizen and government power manifests in conflict between popular
sovereignty and constitutional democracy (Zaslove 2008). Discontent between voters’ ideals
of their influence can conflict with perceived responsiveness of government and the quality
of available avenues. This contrast takes the form of democratic deficits. As individuals do
not have direct control over government, political parties are a tool for citizens to aggregate
common interests and hold greater influence over government (Michels 1915); however, the
increased power of collaboration diminishes the control of the individual as the organization
grows. Conflicts can arise through different groups—represented through parties—struggle for
hegemonic control, or through a party’s paradoxical nature to act in its own interests (Mudge
& Chen 2014). In return for increased organizational strength, individuals concede a certain
amount of their individual political efficacy. Voters and parties engage in a trade–off between
ones’ influence over interest articulation and the strength of the party, increased with larger
membership that decrease individuals’ influence. Deficits arises when the balance between
individual influence and party autonomy moves out of a voter’s ideal range, in which the agent is
not acting in their favor. This can lead one to feel that they have granted too much influence to a
group that gives too little in return. Perception of individual powerlessness can be tolerable when
the organization and a supporter hold the same interests; however, an increasing disconnect
between preferences makes one’s inability to influence the movement more poignant. Parties
fulfill a mediator–like role between citizen and state (Katz & Mair 1995, Michels 1915). It is
this intermediary role that shifts the tension of responsiveness and organizational agency from
a conflict between the government and citizens to tensions between individuals and parties.
Inefficacy and disaffect arising from the sense of democratic deficits therefore manifests in an
individual’s alienation from a political party. Where the traditional, mainstream parties are
the dominant force in government and form the governing coalitions, one can become alienated
from the establishment. A sense of political powerlessness within the mainstream parties acts
as an alienating force.
When selecting a party, ideological proximity is not individuals’ only concern. A
party needs to have the trust of voters and the capability to enact legislation (Labzina &
Schofield 2015). The participation decision is based on the belief that action will bring about
the desired policy, while political decisions and party selection is based on the who will grant
the greatest level of influence; a party’s supporters want an organization in which the balance of
19
power is in their favor. In the formation of, or in supporting a political party, individuals seek
greater power in government in order to implement desired policy. Aggregated interests of like–
minded voters increases influence, alleviating powerlessness alienation (Michels 1915). Through
collaboration, one seeks to minimize “the expectancy or probability held by the individual that
his own behavior cannot determine the occurrence of the outcomes, or reinforcements, he seeks”
(Seeman 1959, 784). However, if unable to influence the political organization, the voter will
again face powerlessness. A party that is no longer a representation of the supporter’s will defeats the organization’s purpose, leading the support to another organization that upholds their
will. As for the tension between the general party membership and its leadership, one may also
face alienation from the perception that the party elite are unresponsive to, or unrepresentative
of the institution’s base. Likewise, the organization may begin seeking its own survival over the
interests of its supporters, colluding with other parties to form a cartel–like, oligarchic organization (Katz & Mair 1995, Michels 1915). Manifested within the governmental institutions,
a cartelized party does not require the support of its constituency, using public resources and
colluding to reduce competition to solidify their power. One will not support a party that has
alienated them, and will alternatively seek organizations in which they have greater influence.
Voters want a movement that can drive change, and they want to be able to control the direction. Inherent struggle between individuals’ influence and parties’ power establishes dynamics
of contention (Mudge & Chen 2014, Oberschall 1978). But when will flash–points emerge, and
what are the effects of voters’ behaviors1 ?
3.2
Social Movements and Party Choice
Work on protest participation and non–institutional action provides insight for voter–party relations and mainstream versus outsider support. While voting behavior and social movements
are for a large part are separate literatures, treated as distinct (McAdam & Tarrow 2010), there
is a connection between them: origins and behaviors of non–institutional, collective action and
institutional, individual actions share common roots and goals. External efficacy has implications for participation, both focused within and outside government, and its effects extend
into party choice. Perceptions of government responsiveness forms the connection between so1
While the institutional reaction to the contention is of equal interest, this study focuses on the individual–level
behaviors of voters, and not party strategy
20
cial movements and voting behaviors, serving as the common denominator. Treating behaviors
within and outside institutions as separate phenomena hinders the understanding of individual
action (Seligson 1980), suggesting that the basis for the two are equally separate. As an attitudinal foundation for social movements, the impacts of efficacy will be similar for institutional
actions. The attitudinal relationship between government and citizens, based in efficacy and
satisfaction, underly the dynamics of contention. Alienation and affective motivations serve to
bring the two forms of political action under a common repertoire, in which unconventional
behaviors, either expressed through governmental means or outside them, will have the same
motivational factors and goals. When individuals feel they have lost control of government, that
the norms of democratic governance have been violated, one’s considerations pivot away from
ideology and shift towards re–exerting influence within government. Where the inefficacious
engage in protest to exert influence, the alienated voter turn to unconventional voting behaviors to have their voices heard. Normative attitudes towards the roles of citizen and political
parties within democratic society supplants ideological considerations when the voter perceives
the parties’ power too strong, the organization cartelized, or serving its own interests (Katz &
Mair 1995, Michels 1915). A perceived power imbalance leads to voting behaviors intended to
re–exert influence, much the same as political protest.
Gamson’s theory in Power and Discontent concerns non–traditional political actions,
the roots of which are in the power struggle and satisfaction (1968). Individuals who are
sufficiently dissatisfied, externally inefficacious, and internally efficacious2 will engage in non–
institutional action, such as protest (Craig 1980, Craig & Maggiotto 1981). The hypothesis
complements previous work on political action—in particular, effects of efficacy on turnout—but
purposes an alternative, in that alienation will not necessarily lead to withdrawal from politics.
Instead, the politically powerless, unable to influence the government within established means,
will use more contentious actions outside its institutions to exert influence. Treating efficacy as a
resource to overcome the costs of participation (Sigelman & Feldman 1983, Valentino et al. 2009)
implies that the inefficacious are unlikely to engage in political action; however, individuals
motivated by expression use protest as a means to voice grievances against a system that has
alienated them (Klandermans 1996). Individuals can posses attitudes on the effectiveness of
institutional and non–institutional action separately, with inefficacy in the former and efficacy in
2
Internal efficacy is not examined in this project. See Appendix A.
21
the latter promoting non–conventional action (Wolfsfeld 1985, Wolfsfeld 1986). While Gamson’s
hypothesis—and Craig’s reformulation—has received empirical support in both institutional and
non–institutional action (Craig & Maggiotto 1981, Nachmias 1974, Paige 1971, Pollock III 1983,
Seligson 1980, Wolfsfeld 1986), the work has not been applied to explain unconventional voting
behaviors. The two phenomenon are not evaluated under the same framework, which this paper
serves to test the implications of Gamson’s hypothesis for vote choice. Where unconventional
action outside government’s institutions can be motivated by external inefficacy and disaffection,
the attitudinal components will be influential for non–traditional party selection.
Within the repertoire of unconventional behaviors, one may choose to act outside institutions or within the electoral process. Faced with alienation, the individual may withdrawal
from the electoral process, electing to use non–institutional behavior—such as those in hypothesized by Gamson. However, before dropping out of the electoral process, one will seek influence
directly with system–focused expressions. Previous work has shown how powerlessness, mistrust, and inefficacy can depress participation; however, these political attitudes, and inefficacy
in particular, do not inherently lead to withdrawal. The politically disinterested and disengaged
will selectively drop out of the political system at lower levels of disaffection. Their withdrawal
is more a product of their preexisting alienation from the little amount of utility they place
on political participation. Likewise, those placing little value in political action are unlikely
to engage in unconventional behaviors. But those alienated not by selection, but through decreased influence, those whose preferences fall on deaf ears, are more resistant to withdrawal.
Those motivated by expressive desires, ideological preferences, or avocation of economically
advantageous policies, act to change the alienating force (Nachmias 1974).
Given a history of participation, efficacy’s effects on developing a habitual pattern on
voting will play a role in persisting participation. Even through one may feel unable to influence,
they will still turnout (Valentino et al. 2009). Continued failed action will manifest in inefficacious attitudes, increasing the possibility of removing oneself from the electorate and using other
means, but withdrawal is not the first result. In the unconventional behavior repertoire, an individual will first exhaust institutionally–focused participation before growing alienation moves
them to extra–system action. Influencing government through the electoral system becomes
impossible for the individual; political parties are too distant from the voter for meaningful
influence, and the government appears to act without citizen consent. When the system is
22
perceived to use its autonomy too much, disregarding individual preferences, means outside
the system become more attractive. Power dynamics between the state and the individual are
seen to be moving too much in the government’s favor. The relationship between voters and
representatives is too frayed, causing one to place little value on preserving an ineffectual relationship with government authorities, so individuals must exert influence from alternate means
(Gamson 1968). Severed links between the populace and political organizations can serve to
increase alienation and spurs distrust among the electorate (Barr 2009, Keman 2014), where
the distrust can have a spiraling effect, further increasing animosity towards representatives
and depressing participation (Grönlund & Setälä 2007, Hooghe & Dassonneville 2014, Muller
& Jukam 1977). But before an individual decides to withdrawal, they use restorative actions
available within the electoral process.
3.3
Vote Switching
When an individual perceives that their party is consolidating its power and removing citizens’
influence, they will attempt to resolve the power imbalance. Perceptions of the cartelization of
a party, or a growth in oligarchic tendencies, will be met with a reaction by party supports.
However, when faced with decreasing political agency and distrust of authorities, protest is
not the first resort. Voting provides a less costly path for influence as compared to the resources required for demonstrations and other forms of non–conventional political action; one
may individually attempt to restore their influence through the ballot, without problems regarding collective action or fraying their relationship with government authorities (Gamson 1968).
Where a party is an aggregation of interests, uncoordinated efforts of disaffected, alienated
individuals changing loyalty has the effects of their combined influence. A party will lose organizational strength and power within government equal to its number of defectors. When
assuming parties are rationally bounded, having imperfect information of voters’ preferences,
the organizations will alter their positions when learning more of the electorate (Kollman, Miller
& Page 1992). Without expending the resources required for a political movement, the collection of voters leaving the party will have the same effects as an organized movement. Seeing
a loss in support, the party will adjust their policy position to one more closely aligned to a
majority—or plurality—of voters, using voters’ changing loyalties to inform the ways in which
they adjust their stance (Shotts 2006). Where individuals make signaling assumptions about
23
parties’ responses to loss of vote share, as well as gauge the behaviors of other voters (Kselman
& Niou 2011, Myatt 2015), party defection serves as a quasi–social movement: there is no direct
collective action, but the individual assumes others who are similarly–motivated will behave in
a similar fashion as themselves to achieve the same goals. If a party wins, they will engage in
satisficing behaviors, maintaining their position; however, if the election is lost, the party will
change its stances before the next poll (Bendor, Mookherjee, Ray et al. 2006). The size of voter
defections communicates the extent to which a party’s members are disaffected, influencing the
size of the party’s potential adjustments. When one anticipates that enough of like–minded
party supporters will defect, then the possibility of shifting the party’s position is greater.
Vote signaling can be seen as a special case of vote switching, in which one has more
clearly–defined goals. The individual has a more specific message and desired result when
engaging in strategic signaling, as opposed to registering general disaffection. In showing a
party their preferences, voters issue a warning meant to re–exert their influence and make
the organization more attentive to their voice. Through their signal, one may show support
for a specific policy of a niche movement they feel has been neglected (Franklin et al. 1994).
In signaling, voters abandon instrumental considerations in favor of expressive motivations,
knowing that their vote will not likely influence the election results, but that their vote will alter
future behavior of parties. Such a case can be seen in the Swedish Pirate Party, whose voters
were inspired by their support for their uniquely–captured policy space (Erlingsson & Persson
2011). When individuals’ political action is successful and their party adjusts its position, the
individual’s efficacy increases (Finkel 1985). After successfully signaling, voters will return to
their original party. As their alienation was, at least in part, addressed, the individual does not
continue to engage in untraditional voting behaviors, instead returning to traditional spatial
voting.
Less clearly–defined, one’s vote can be used as a soft—or low–cost, more individualistic—
route to register their desires. Casting a ballot behaves as a quasi–movement. While non–
collective and separate from a social movement, unconventional action within and outside the
system are intertwined: institutional forms of the unconventional repertoire are the first resort. Directly registering preferences through selecting government representatives offers the
most direct form of expression by providing a mandate to elected officials; and, alternative to
supporting a party, signaling behaviors can have the effect of removing an electoral mandate,
24
shaping an organization’s future action. Before one will engage in demonstrations, they will exhaust lower–cost institutional actions that serve as a less overtly–contentious means of influence.
Political parties can form a vehicle for protest, functioning as an aggregation of disaffected citizens, giving their discontent greater influence (Miller 1974). Besides cost–benefit calculations,
the habit–forming qualities of efficacy on voting (Gerber, Green & Shachar 2003, Valentino
et al. 2009) elongates the time before which one withdrawals. The more efficacious the individual was before loosing their sense of government responsiveness, the longer it will be until
they resort to extra–system routes. Concerned with social norms and pressures, one may also
use the vote as a socially–approved route to influence. Participation in democratic governance
fulfills one’s sense of civic duty, influencing individuals to varying degrees. Unwilling to engage
in public action such as demonstrating, one may use the vote as an anonymous contentious
behavior. Such unconventional behaviors are indicative of future non–traditional, extra–system
action: political unrest in a given country can be predicted by the extent to which blank and null
votes are cast (Superti 2014). Individuals do not want to increase their alienation from government authorities, and are therefore resistant to undertake provocative actions. Confrontational
behaviors and extra–system action will not be utilized until conventional actions have been
ineffective (Gamson 1968). Faced with continued unsuccessful attempts to rectify alienation
within the confines of government, an individual can progress to using non–institutional actions
from the repertoire of unconventional action.
Wide–scale displeasure of citizens who feel that they have lost all influence within
government makes powerlessness a nationally salient attitude. When individuals become more
alienated and feel that they lack agency in political affairs, their propensity to hold negative
attitudes increases and the individual becomes more oppositional (Thompson & Horton 1960).
These sentiments of parties’ failure to uphold their representative duties provides opportunities
for outsider movements to break through, contributing to the fading mainstream monopoly
over electoral politics (Dalton 2000, Ignazi 1996, Keman 2014). Using us–vs–them appeals or
anything–but–the–mainstream rhetoric makes these movements a more attractive option to the
alienated portion of the electorate (Barr 2009, Levin & Eden 1962, Miller 1974, Thompson &
Horton 1960, Zaslove 2008). A signaling option or a movement that will grant an individual
greater influence serves as a viable alternative to those feeling politically powerless, who perceive
that their current party has taken too much power. Given the presence of an outsider movement,
25
the party serves as an “exit option” for distrustful voters, in that an individual is given an
alternative outside their current party or establishment politics; and in the absence of such a
group, a voter will remain with their current party (Hooghe et al. 2011).
The question arises “given differences in policy preferences or ideology between a voter
and their party, at what point will the voter change allegiance?” Values shifts and the decline of
political cleavages, described through post–materialism (Inglehart & Flanagan 1987) or ”New
Politics” (Conradt & Dalton 1988) describe an environment in which differences between individuals and parties are more likely to be salient. The emergence of new policy areas and
considerations of non–materialist goals opens the possibility for greater differences between traditional mass, catch–all parties and the electorate. But when will an individual, concerned with
differences in their own and their party’s beliefs, switch to a newly–emerging movement, such
as the greens? Given that two individuals share a similar ideological profile, policy preference
is unable to discern which will change their party loyalty and which will not; however, the
relative efficaciousness of the two voters can inform their propensities to defect. The voter who
is dissatisfied with their current party, feels they lost their influence, or desires for the party to
adopt the policy position of another will be more likely to cast their ballot for an alternative
organization. When party members feel that they have control over the movement, there is
no need to change allegiance. Given influence within a party, one is more able to dictate the
organization’s proposed answers to the new issue areas. The decision to switch parties arises
from the individual’s sense that they do not have this influence, the ability to shift the party.
Dissatisfied with the party acting independently of the voter, the previously–loyal support will
seek another. Once one becomes sufficiently dissatisfied or alienated from the organization,
they will change their support to a party that is more in line with their ideology, will give them
greater influence in organizational decision–making, and more purely expresses their desires.
H1 The inefficacious are more likely to change which party they support than the efficacious.
H2 Those that are dissatisfied with government are more likely to switch their vote than the
satisfied.
26
3.4
Mainstream versus Outsider Support
While elections have seen increasing volatility, support for the mainstream, establishment parties has also been decreasing. What factors influence an individual to support outsider political
parties? Social movements literature informs the decision to engage in non–traditional and
non–institutional action, but will support unconventional political parties have similar motivations? As a deviation from the status quo, casting a ballot for an outsider political party can
be evaluated as a form of non–traditional voting behavior. When an individual is alienated not
only from a single party, but from politics–as–usual, the individual will follow the repertoire
of unconventional political behavior, becoming more prone to support organization challenging
mainstream, establishment dominance. No longer able to influence government through standard electoral choices, the inefficacious voter’s decision–making changes. Transitioning from
ideological representation concerns, the individual begins to focus more on which organization
will grant them the greatest control, rectifying their alienation. As opposed to considering
the power relationship between the self and an individual party and its leadership, the power
dynamics between oneself and the political elites becomes more salient; concerns of influence
maximization moves from one’s control within their party to a more broad evaluations of the
role of their voice within government more generally. In such situations, perceptions of democratic deficits motivate the individual to take unconventional actions to restore the norms of
democratic governance.
While there is an ideological component to outsider political support, there is a potent
effect of alienation, separate from liberal–conservative positioning and policy preferences. The
rise of anti politics and outsider parties can be seen as a social movement in its own right:
individuals, while not acting collectively, will use means from the unconventional repertoire
of action to restore their influence within government. The vote is an individualistic form of
action, however, electoral decision–making will be influenced by similar factors as participation
in a movement; although acting in different arenas, the two phenomenon share common goals
and motivations. There is an element to non–traditional voting behaviors that is separate
from ideology, in that the party decision is motivated by an individuals attitudes towards
government and their perception of the relationship between the elected officials and their
constituents. Belief in the cartelization of the mainstream and the dominance of an oligarchic
political class out of touch with its citizenry spurs a reaction by voters. In this instance, the
27
exit option is not necessarily a particular political party, but any outsider, non–mainstream
party. Given the choice between multiple non–establishment alternatives, ideological factors
and spatial considerations will take over, but the decision to support an outsider party is not
based in ideology. Declining mainstream party support and the surge of untraditional politics
cannot be explained within the context of shifting ideologies and new political orientations;
alternatively, non–traditional voting behaviors are more in line with a social movement: the
reaction of voters to alienation from the status quo. There is an action and reaction between
voters and parties in the struggle over political power.
Corresponding with political powerlessness, the growth of outsider parties can be motivated through an affective response towards the establishment. A strong “throw the bums
out” sentiment from inefficacy can lead to increasing anti votes (Levin & Eden 1962), in which
an individual is not issuing support for a party, but registering a dislike of the alternatives.
Restorative, affective actions do not suggest a positive evaluation of the chosen party, but instead act as a display of the lack of support for the mainstream and current state of political
affairs. As with the power dynamics, affective voting is not ideologically–based, alternatively
motivated by disillusionment with a goal of registering discontent. Individuals express exasperation with their total loss of control over government and the party system; the disaffected
voter mobilizes their discontent in non–conventional voting, much like how dissatisfaction mobilizes into non–traditional extra–system action. Given alternatives, an anti–establishment or
general outsider party provides an exit option from the oppressive traditional politics (Hooghe
et al. 2011, Hooghe & Dassonneville 2014). The sense of powerlessness accompanied by perceived threats from disadvantageous policy drives individuals to action in order to stop the
threat (Valentino et al. 2009). Coinciding with this effect, the inefficacious are more likely to
hold negative attitudes (Thompson & Horton 1960).
This affective response can be seen during European Parliament elections. Where some
voters see that these elections will not have direct effects on their lives, they present a near
costless opportunity for expression. When dissatisfied with their chosen domestic organization,
individuals are more likely to use the European Parliament punitively (Hix & Marsh 2007, Spoon
& Hobolt 2010, Van der Eijk, Franklin & Marsh 1996). Diminished strategic concerns create
perceptions that vote wasting at the supranational level is lower cost, removing instramental
concerns and allowing the individual to act in a more expressive fashion. In the case of the
28
disaffected voter, expression takes the form of discontent. Accompanying disaffect with respect
to a specific party, the European Parliament elections may also be used as a more general
expression of dissatisfaction, in which the individual does not send a message to a single party,
but the establishment more broadly ((Treib 2014). From a similar perspective, a distrustful
domestic electorate is more likely to select outsider or protest parties (Hooghe & Dassonneville
2014, Hooghe & Marien 2013). However, in the non–conventional action context, distrust is not
enough: efficacy is necessary to explain non–traditional behaviors (Pollock III 1983). Voters
become disgruntled when they do not feel that they are represented. When the party is perceived
to abandon its constituency, the voters have an affective, non–ideologically driven response.
This discontent can be an important component in an electoral system. Non–traditional voting
behaviors can lead to re–alignment with voters, in which parties re–adjustment in response
to voters’ actions (Miller 1974); alternatively, there can be a spiral of distrust (Hooghe &
Dassonneville 2014). The declining mainstream support is a social movement–like phenomena,
with the inefficacious and dissatisfied following the unconventional behavior repertoire. Support
for new, outsider parties has roots in non–ideological factors, in which an individual does not
have strict spatial or policy goals; instead, the goals are focused on restoration of positive affect
and political influence.
H3 Inefficacious voters are more likely to support a non–mainstream political party than the
efficacious.
H4 Individuals dissatisfied with government are less likely to vote for a mainstream party
than the satisfied.
3.5
The Role of Ideology
This is not to say the ideology is an unimportant factor in electoral choice: instrumental factors
and voters’ proximity to alternatives is a primary motivation behind party choice. However,
individuals behave spatially in a normal electoral context when engaging in conventional voting,
but the inefficacious and disaffected portions of the electorate do not base part choice on conventional motivations. Political powerlessness and disillusionment promotes behavior according
to the unconventional repertoire, potentially resulting in spatial incorrect voting. Research on
party choice has examined a two–step process, which can be interpreted through an attitudinal
29
understanding. Voters first evaluate a party as a standard alternative or a protest option, and
then vote spatially or non–ideologically, respectively (Van der Brug, Fennema & Tillie 2005);
an efficacy–ideology provides a similar two–step evaluation. As opposed to deciding whether a
party is normal or a protest option, the voter considers which party will maximize their influence: the fist stage in the processes is the decision to engage in traditional of unconventional
behavior, where alienation results in the latter. After deciding on a course of action, the individual will then choose a specific party spatially. While work on anti–immigrant parties have
found that they are treated as standard parties (Van der Brug et al. 2005)—not serving a protest
function—examination of electoral behavior and social movements as equivalent phenomenon
in different contexts may result in a alternate conclusion. A given voter using unconventional
voting will select an anti–immigrant party when it is the most closely aligned alternative, but
that does not indicate that an anti–immigrant party is treated as a traditional organization.
When voters’ perception towards government sours, democratic deficits, and overall satisfaction with politics deviates from acceptable levels, ideology alone is not sufficient.
When there is not a substantial inefficacious base of voters, spatial modeling can explain party
choice, as the majority of individuals engage in standard ideology–based decisions. But, when
a large number of voters engage in unconventional behavior, a spatial understanding alone
cannot explain voting that is not strictly ideological. Individuals’ sense of political empowerment and satisfaction provide an explanation for the voting that cannot be captured through
liberal–conservative party evaluations. Efficacy provides the necessary condition for “normal”—
ideologically–based—political behavior. The belief that an individual’s political action will have
the intended effects is a prerequisite for actions motivated by instrumental concerns. When inefficacious, individuals will first select the movement that will grant them the greatest influence,
or send the most effective signal to their most preferred party. If there are multiple exit options, one will choose the alternative most closely aligned to their preferences. The importance
of attitudes towards the system and its representatives is deciding which type of party one will
support. However, efficacy and satisfaction will have less effect on specific party choice within a
party family. With a variety of mainstream or outsider movements to select from, after deciding
which type to support, the individual will then use ideological proximity to select which party
to support.
H5 Ideological and policy preferences will be greater factors in selecting a specific party, while
30
efficacy and satisfaction will be less influential in the decision of which party to support
within the mainstream or outsider categorizations.
Instrumental considerations alone are insufficient for explaining vote choice. Incorporating expressive desires provides additional insight into behaviors and increases explanatory
power. Work on vote choice has begun to examine more than spatial concerns with socio–spatial
models; however, individuals’ efficacy, their perceptions of democratic deficits, their relationship
with government, and their ideal level of influence has been underutilized. The vote is more than
a means to register desired policy. Individuals can use the ballot strategically, sending signals
in an attempt to reform their parties future actions or voice discontent, as well as select the
movement that will give the individual that greatest amount of control over government. Motivations behind participation and political action are treated as separate from political choices;
the factors that push individuals to the polls are underused in research on party choice. There
is a disconnect between the work on institutional and extra–system action, and the two are
treated as if independent from one another. Alternatively, this project proposes a social movement evaluation of unconventional voting behaviors. Electoral behaviors and social movements
are common phenomenon with a repertoire of unconventional actions, but in different contexts:
the former is institutionally–focused and the latter works outside government. Before an individual uses non–institutional restorative actions, they will engage in the political process and
exhaust existing, established means to exert influence. There are two motivations for restorative
actions: efficacy and the power–dynamics between citizen and party, and the affective response
to perceived violation of norms sourced from government dissatisfaction.
4
4.1
Data and Methods
The Netherlands
The three main organizations in Dutch politics are the center–left Labor Party, Partij van de
Arbeid (PvdA); the Christian Democratic Appeal, Christen–Democratishe Appel (CDA), a
standard Christian Democracy party; and the Liberal center–right People’s Party for Freedom
and Democracy, Volkspartij voor Vrijheid en Democratie (VVD). PvdA and VVD were founded
1946 and 1948, respectively, while the CDA emerged in 1977 after the fusion of three Christian–
31
Democrat parties: the Catholic People’s Party (KVP), the Anti–Revolutionary Party (ARP),
and the Christian Historical Union (CHU). These three organizations have dominated post–
industrial Dutch politics, with one, two, or all three comprising every coalition since 1971;
The CDA has been a member of the most, then the VVD—who is often in government with
CDA—and PvdA the least3 .
The Calvinist Reformed Political Party (SGP) is currently the oldest party in the
Netherlands, with an old–right religious ideology and has never entered into a governing coalition. The Socialist Party (SP) was founded in 1971 and has had moderate electoral success, but
the party has always remained in opposition. Movements that have emerged more recently are
much more heterogeneous and less traditional. De–pillarization of Dutch politics—the erosion
of class and religion–based voting, accompanied by general secularization—have shifted the electoral dynamics, as has growing post–materialist values (Van Holsteyn & Irwin 2003, Inglehart
1981, Inglehart & Flanagan 1987). Compounding the changes, the 1971 elections were the first
in which the voting age was decreased from 21 to 18, enfranchising a younger electorate. Desire for greater democratization and an American–style presidential democracy gave rise to the
radical–democrat movement: the Democraten ’66 (D66), named after their ideology and year
of formation. D66 has been the most successful outsider movement, and has entered into four
governments. Environmentalist parties unified into the GroenLinks (GL) in 1989, or the Green
Left. Although never entering into a national coalition, the movement has been quit successful
in provincial elections. The D66 and GL are representative of the beginning of a new type of
Dutch parties that continued to emerge.
Beginning in the early 2000s, a variety of new movements formed and began to grow.
An outsider Christian, more centric alternative to the SGP, gained prominence when the Reformed Political Alliance (GPV) and Reformatory Political Federation (RPF) merged into the
Christian Union (CU) in 2001. After electoral success in 2006, CU was invited by CDA to join
their coalition with PvdA. Occupying a more unique ideology space, the Party for the Animals
(PvdD) was founded in 2002 as a testimonial party for animal rights and welfare, winning several
seats in parliament after the 2012 elections. Perhaps the most defining Dutch political events
of the early 2000s was the quick ascension and collapse of Pim Fortuyn List (LPF). Expelled
3
Given the electoral preeminence of the PvdA, CDA, and VVD, these three organization are classified as the
mainstream, establishment parties.
32
from Livable Netherlands in February 2002, Pim Fortuyn began his own list of candidates for
parliament. Outside traditional ideological definitions, “Fortuynism” was anti–elite, populist,
anti–immigrant, but supported traditional Dutch social liberalism. In May 2002, Pim Fortuyn
was assassinated a week before the general elections, in which LPF won the second most votes
and gained 26 seats. The party joined a coalition with CDA and VVD, but the government
collapsed within a few months. Elections were held January 2003, where LPF only held onto
eight seats. LPF won no seats in the 2006 election, and the movement dissolved 2008. In a sort
of continuation of anti–immigrant politics, Geert Wilders left the VVD in 2004, incorporating
the Party for Freedom, Partij voor de Vrijkeid (PVV), in 2006. After the 2010 elections, PVV
agreed to support the VVD in its coalition with the CDA.
4.2
Data
To test the social movements approach of an attitudinal–ideological basis of vote choice, I
use the Dutch parliamentary Elections Studies (DEPS) integrated data set, which contains
data on the 1971, 1972, 1977, 1981, 1982, 1986, 1989, 2002, 2003, and 2006 elections. The
Netherlands institutional and political landscape are an optimal environment for analysis of
individual–level factors affecting voting behavior. There are no electoral districts in national
elections for the 150–member parliament; the entire country acts as a single district, voting
for all MPs simultaneously. The single–district nature creates a highly proportional system in
which a political party only needs to gain 0.66% of the vote in order to win representation.
High proportionality reduces strategic voting incentives that create concerns of vote wasting
(Franklin et al. 1994), which allows for a wide range of alternatives and reduces the instramental–
expressive voting competition. The low threshold permits smaller changes in factors influencing
vote choice to manifest in electoral results, permitting effects from changes in voters’ behavior
to be readily examined. In less proportional systems, the effect of a single factor on vote
choice must be great enough to overcome strategic considerations, which are reduced in the
Netherlands, allowing for expressive motivations to be more prominent. Dutch politics face a
similar environment as other European parliamentary democracies, but more uniquely offers
the opportunity to measure these factors’ effects on voting.
33
4.2.1
Dependent Variables
The DPES includes a question on voters’ intended party choice in the upcoming general election4 . Respondents’ previous party choice was included in the survey, as well as which party
the individual actually voted for, which is asked in the post–election waves. The parties included are: PvdA, CDA, VVD, D66, GL, SGP, CU, LPF, SP, KVP, CHU, ARP, GPV, RPF,
Centrum–Demokraterne, and “Other.” For the 1971, 1972, and 1977 studies, the ARP, KVP,
and CHU are coded as CDA. Although the Centrum–Demokraterne (CD) are included among
the party choices, only 12 respondents indicated that they would support the party. Due to
the low level of indicated support, CD is included in “Other.” The “other” category creates
an issue with analysis of party choice due to the heterogeneity of the movements included. For
example, both the PvdD and PVV are included in “Other.” The category prevents inference of
theses specific movements’ support and the ideological influences behind voting for an “Other”
party; however, the category still allows analysis of insider vs. outsider support. This DPES
question was used to form the categorical vote intention, previous party supported, and current
party supported variables. For analysis of vote intention, “undecided” serves are the reference
category, and individuals who did not intend to vote are excluded.
The vote switching variables is dichotomous, with the responded being coded as one
if the party they voted for previously is different from the party they intend to vote for in
the upcoming elections. Although the respondents’ self–reported past voting—up to five years
prior—may not be the most reliable, for all intents and purposes of this project, the factors
that contribute to an individual either misremembering or incorrectly reporting their choice is
of equal interest to the actual voting record. Those that either did not previously vote or do not
intend to vote in the current elections are included in the variable. The inclusion of non–voters
allows for analysis of the voters leaving or re–entering the electorate. Undecided respondents
are not included in the variable.
Intention to vote for a mainstream party is a dichotomous variable, coded as one if
the responded will vote for PvdA, CDA, or VVD. For this paper, mainstream versus outsider is
defined in terms of participation in governing coalitions, as opposed to ideological or other classifications; however, the three parties classified as mainstream are relativly centrist. Frequent
4
The question was not included in the 1972, 1982, and 2003 studies
34
participation in the government suggests broad support for a party and its establishment status.
PvdA, CDA, and VVD are classified as mainstream, establishment organization because they
have received the top vote shares in all elections since 1971, excluding 2003 when the LPF took
second, and PvdA received the fourth most votes. Although other parties have participated in
governing coalitions, their presence is less consistent. Only those that intend to vote and have
decided who they will support are considered.
4.2.2
Independent Variables
An efficacy variable is created from yes/no answers to the external political efficacy items included in the integrated data file, in which agreement with the question indicates efficaciousness.
The number of question asked ranges from three to five, with the later surveys including more
questions than the earlier. Question wording reflects the operationalization of efficacy validated
in the ANES (Niemi, Craig & Mattei 1991). The number of “yes” answers—coded as one—
is divided by the total number of questions the respondent answered, as to create a variable
consistent across surveys in which the range of efficacy scores is between zero and one.
For government satisfaction, the provided DPES variable5 is used. Respondents select
from a five–item scale from one being “very satisfied” to five being “very unsatisfied.” Coded in
the DPES as a high number equating to more dissatisfaction, the variable was switched for the
analysis: a score of one represents most dissatisfied, with a higher score indicating a higher level
of satisfaction. The question regards general satisfaction with the government’s performance
over the previous years, as opposed to policy or institutional dissatisfaction. General satisfaction
with the government allows respondents to broadly express incumbent affect, as opposed to
restricting discontent to a single factor, such as policy. It is assumed that dissatisfaction with
the functioning of democracy is encapsulated in the efficacy variable, as a more inefficacious
individual will express frustration with their lack of influence; therefore use of satisfaction with
democracy would create redundancy in the analysis.
A policy preferences item is created in a similar fashion to efficacy. Like the efficacy
questions, the number and issue items have some variability by study. The variable is constructed by how respondents place themselves on a scale, from one to seven, regarding various
5
Not included in the 1982 study
35
issue areas. Their answers are summed and then divided by the number of the questions answered. The variable represents respondents’ average position within the issue space. A one
represents the traditionally more liberal policy position, with seven the conservative position.
This construction allows for a consistent variable across all study years, as well as increasing the
number of complete cases. If a respondent refuses to answer a specific question(s), but does give
their opinion on other areas, then their policy preferences are still recorded. Due to the lack of
data on respondents’ perceptions of parties’ issue stances, and to keep the model parsimonious,
the difference between the respondent and political parties on issues is not included. Based on
previous work on spatial voting behaviors, it is assumed that if policy preference is a significant
factor in vote choice, then it is because of the individual’s and the alternatives’ relative positions
on the issue items (Downs 1957, Van der Brug & Fennema 2007, Quinn et al. 1999).
A left–right self–placement is provided by the DPES in the pre–election wave studies6 .
Whereas policy preferences regards specific issue areas, the left–right scale is more general
measure of a respondent’s ideology. Respondents select a position from zero, the most liberal,
to the most conservative, 10. As with policy preference, the significance of respondents left–right
placement is assumed to be based on proximity to parties7 (Van Der Eijk, Irwin & Niemöller
1986).
The control variables are: income, self–image of one’s social class, the highest level of
education attained, and age. All the variables are constructed such that an increase indicates
a higher level of the respective characteristic; if a variable was categorical, it was converted to
a numeric scale. While traditional cleavages are less prominent in voting behavior (Inglehart
1981, Inglehart & Flanagan 1987, Van Holsteyn & Irwin 2003), they are still expected to have
some influence over party choice—especially with inclusion of earlier elections. Self–description
of class asks respondents to place themselves within one of the given classes, while income
asks net annual household earnings and categorizes respondents into one of twelve quantiles.
From work on radical–right and anti–immigrant parties, it is expected that higher income
and self–perceived social class will be associated with mainstream support (Van der Brug &
Fennema 2007, Arzheimer 2009). Due to privacy regulation, respondents’ exact age are not
included in all studies; therefore, the age variable represents categories of approximately five
6
Left–right self–placement was not asked in the pre–election wave of 2006, so the post–election wave scale was
used
7
Figure 3 depicts the correlation between a voter’s left–right placement and the position of their chosen party
36
years each, ranging from 17 to 99 years old.
4.2.3
Variables Across Elections
Although the DPES does not track the same
individuals across elections, results from the
Figure 2: Attitudes and Behavior Across Time
various studies are able to depict general
trends in Dutch political life. The Nether-
Voting, Efficacy, and Satisfaction
1.0
lands ended compulsory voting in 1967, but
0.8
has maintained relativly stable voter turnout
0.6
at about 80% of those registered casting a balPercent
lot. While turnout has been decreasing across
0.4
Europe, decreases in Dutch participation have
0.2
not been as severe. Results from the 12 electoral studies from the 1971 to 2006 national
0.0
Turnout
Mainstream Vote
Efficacy
Satisfaction
1970
1975
1980
1985
1990
1995
2000
elections appear in Figure 28 .
ing, Efficacy, and Satisfaction” plot traces re-
Year
spondents’ turnout in the current election,
Electoral Volatility
the proportion supporting mainstream par-
Total Vote Switching
Switch within Mainstream
Switch to Mainstream
Switch to Outsider
Switch within Outsider
0.4
The “Vot-
2005
ties, as well as their efficacy and government
●
creased since the late 70s; however, the 1982,
0.2
1986, and 2002 elections saw sizable increases.
●
While the proportion of those indicating they
●
●
voted has remained within a range of eight
0.1
Percent
0.3
satisfaction score. Overall, turnout has de-
●
0.0
percent, efficacy and satisfaction have been
1970
1975
1980
1985
1990
1995
2000
2005
more variable across studies.
Year
Voter turnout percentages are based on respondents self–reporting to
DPES, not official data. Vote choice was based on the party the
respondent supported in the current election, not intention; vote
switching is also based on previous party supported and the current
party supported.
Respondents
were less efficacious in the 70s, gaining confidence in their ability to influence government
into 1989 election, but once again facing in-
8
Respondents self–reported voting, and not official data, are used. While individuals over–report voting, the
general trend follows that of actual turnout
37
creasing powerlessness in the remaining studies9 . Dissatisfaction is highest for the 2003 respondents, while those surrounding in the 1977 and post–1994 surveys are most content with
government performance. The fortunes of Dutch mainstream parties reflects that of Europe’s
political establishment: their electoral prospects have been decreasing. The numbers indicating
they intended to support either PvdA, CDA, or VVD was a little more than 60% in the 1970s,
but was at its highest levels in the 1980s. However, in the elections proceeding the 1986 polls,
the proportion of mainstream voters has precipitously declined.
Changing dynamics of Dutch elections can be seen in the Electoral Volatility plot.
Total rates of vote switching and changing between and within party types are plotted using
respondents’ answers to current and previous party for which they voted. Following the decline
in the mainstream’s vote share, those indicating that they changed their party support has
increased. Elections in the 70s and 80s had relativly low volatility, but increasing numbers in
following studies began changing party loyalty thereafter. Diminishing party loyalty accompanied the post–1986 mainstream decline, in which vote switching began quickly increasing after
the elections of 1989. Proportion of those supporting a different party was highest in 2002, with
over 40% changing parties; however, this may be exceptional do to the rise of LPF. The second
highest proportion occurred in 200610 at just below 40% casting a ballot for a new party.
4.3
Methodology
For the data analysis, I use Bayesian inference11 . Bayesian statistics differs from frequentist, or
classical approaches in the treatment of uncertainty, as well as in the construction of questions
about probabilities. Bayesian inference quantifies uncertainty through probability densities,
treating uncertainty through a process in which there is a prior belief, evidence is collected, and
then the belief is updated to reflect the new information. The process of Bayesian inference can
9
Extrapolating from the independent surveys, the 2002 election appear to have served a pressure–valve like
function. There was a peak in inefficaciousness leading into balloting that swept LPF to power. The success of
LPF restored some level of efficaciousness, as respondents’ in the 2002 average efficacy score is .457, while those
answer the 2003 DPES have an average score of .393.
10
Data on vote switching for 2003 is not available
11
Use of Bayesian statistics has become increasingly popular in Political Science (e.g. Achen 2006, Bartels 2009,
Bernardo 1984, Bonneau & Cann 2011, Rosas 2006) as well as in the social sciences more broadly (Gill 2014).
38
be represented by Thomas Bayes’ rule, which is as follows (1763):
P (A | B) =
P (B | A)P (A)
P (B)
P (A | B) ∝ P (B | A)P (A)
The posterior, or updated, probability, P (A | B) is proportional to the prior probability, P (A),
times the likelihood of the parameter, given the known information P (B | A). The two only
differ by a constant, which is the total probability of the data across all possible parameter
values. Probabilities of values for a parameter, given the data, is proportional to the likelihood of
the data under the parameter value, multiplied by the prior belief of the value of the parameter.
Alternatively, an initial degree of belief—or ignorance—regarding a hypothesis is combined with
the probability of the observed evidence being generated given that the hypothesis is correct,
resulting in a new state of belief in the hypothesis.
Frequentist statistics, as the name suggests, makes inferences from the relative frequency of events. Data are treated as non–fixed, taken from repeatable, random samples; the
population parameter is treated as fixed across data, with inference of the parameter drawn
from the repeated samples and asymptotic theory. In Bayesian inference, the data are treated
as fixed realizations, generated from the true parameter values. The credibility of estimates
for parameters is assessed on the extent to which they fit the data. Uncertainty regarding
unknown parameters is expressed as a random variable, where values that are better able to
describe the given data are more probable. While the population parameter that created the
data is a single value, the state of knowledge regarding its true value is expressed as a density;
and for different data, there can be different parameter values that generated the observations.
Bayesian statistics makes use of inverse probabilities, focused on questions regarding the cause
of an event, given its effects. Although frequentist and Bayesian statistics vary philosophically
and interpretation of findings, the empirical results produced by the two methods are typically
the same; in regression analysis, the estimates tend to be similar, but the interpretation of
results differs, with frequentist and Bayesian interval estimation differing in implication.
Although named after the English Presbyterian minister, Pierre–Simon LaPlace developed Bayesian statistics for his work on celestial mechanics. LaPlace independently discovered
39
a more generalized version of Bayes’ Rule, depicted in A Philosophical Essay on Probabilities
(2012):
P (B | Ai )
P (Ai )
j∈Ω P (Aj )P (B | Aj )
P (Ai | B) = P
After LaPlace’s developments and application of inverse probabilities for his research, Bayesian
statistics laid mostly dormant until the mid nineteenth century during the second world war.
The computational intensity and often–produced intractable integrals limited the applicability
of Bayes. The advent of computers resolved calculability issues: Markov Chain Monte Carlo
(MCMC) methods implement numerical integration for estimation of high–dimensional integrals that were previously unsolvable. First used in physics, Monte Carlo methods were used
to model stochastic processes, such as particle movements (Metropolis & Ulam 1949). Later
combined with Markov Chains, MCMC was then applied to statistics (For a history of MCMC
methods, see Richey 2010). The result was the Metropolis–Hasting algorithm (see Metropolis,
Rosenbluth, Rosenbluth, Teller & Teller 1953, Hastings 1970). The use of MCMC allowed
for Bayesian statistics to be more readily calculated, granting greater applicability. While the
theory behind Bayesian statistics was debated, applied Bayesian statistics was used to solve
problems that frequentist probability could not. Pertinence of Bayesian methods in applied
problems—such as cracking the Enigma Code—encouraged a renewed interest. Since World
War Two, Bayesian statistics has become more popular in the natural and social sciences,
prompting most frequentist regression models to be adapted into a Bayesian environment.
The most contentious aspect of Bayesian statistics was—and to some degree, still is—
the requirement of a prior: P (Ai ). In order to make inference from inverse probability, an
initial degree of belief—the prior knowledge of the parameter—must be incorporated through a
probability density. Prior information can influence results, although the effect decreases with
sample size and sensitivity of results can be readily tested; however, the possible impacts of
a prior are less than that of the likelihood function, which affects frequentist and Bayesian
inference alike. Although the most disputed (Gelman, Carlin, Stern & Rubin 2014), the prior
can also be an incredibly beneficial component for social science research. As an illustration,
Bayesian statistics and the use of priors can improve polling predictions. By incorporating new
polling data into the previous through Bayesian updating, Bernardo (1984) was able to predict
40
the Spanish national elections with a higher level of precision. Old estimates were used as the
prior for the new opinion polling, and the new information was combined with the previous
polls to create a new posterior probability density. When constructing a prior, any probability
density, including improper densities12 , may be used.
There are traditional sources of the initial state of knowledge, with the prior density
representing the starting belief, which a new analysis may or may not uphold. When nothing
is known of the parameter, i.e., there is no prior belief in the parameter’s value, one may use
an uninformative prior: a diffuse distribution for which no value is substantively more likely
than any other value. The use of non–informative priors is referred to as objective Bayes13 . An
elicited prior incorporates non–statistical issue–area expert contributions into the parameter
estimation (Gill & Walker 2005). For example, prior knowledge gained by testimony from
government authorities or policy specialists can be included in the analysis through the prior
knowledge. The elicited prior may also be based on findings from qualitative study. Likewise,
previous empirical research may be used to construct a power prior, in which the prior degree
of belief is based on the results of the previous quantitative analysis. In particular interest,
existing regression coefficient estimates can be constructed into a prior probability density for
the current analysis. Bayesian and frequentist models may also be combined with a reference
prior, which uses the parameter estimates generated from frequentist models. For example, a
reference prior for regression coefficients can be the defined with the mean set to the MLEs and
the initial precision equivalent to the inverted sample variance–covariance matrix (Bayarri &
Berger 2004, Lecoutre, Lecoutre & Poitevineau 2001).
However, results can be sensitive to prior specification. The relationship between the
posterior and the prior can be seen as a weighted average of a sort between the data and initial
beliefs14 . As N increase, the results are more influenced by the data, with the prior having
little effect. In large–N survey data, such as the DPES, prior specification has increasingly
little influence on posterior estimates. To ensure that results are not substantially influenced
by the prior, a sensitivity analysis, or robustness evaluation, is used—especially with a low N .
Sensitivity analysis, which is the use of different prior distributions, specifications, and analysis
12
A non–traditional or non–finite probability density. For example, a normal density with infinite variance
Developed
by Harold Jeffrey, who created a rule for determine an uninformative prior. Jeffrey’s Prior is:
p
P (θ) = I(θ), where I(θ) is the Fisher Information.
14
When a conjugate prior is used, there is a clear relationship between the prior pseudo–data and posterior.
13
41
of the change in conclusions resulting from changes in initial conditions more fully explores
uncertainty and assists in evaluation of estimates (Berger 1982).
Whether using uninformative priors or constructing an initial degree of belief, Bayesian
statistics’ has a variety of nice properties. The assumptions of Bayesian analysis are more in
line with the way in which social science research is conducted (Gill 2014). Bayesian statistics
provides an opportunity to study singular, non–repeatable events, such as elections. Since
Bayesian methods do not assume infinite repeatability, one may readily make inference about a
one–time event. The unobserved, unknown causes of various electoral behaviors can be inferred
from the observed effects with the use of Bayesian statistics. And, although the estimation does
not assume repetition, Bayesian statistics still maintain the asymptotic properties of frequentist
regression, while allowing analysis of small–sample data. When there are not enough data for
the limit theorems to apply, the prior still allows for inference. Treating uncertainty surrounding
parameters as a probability density, along with the variety of possible priors, makes Bayesian
analysis more versatile and more widely applicable. While Bayesian probability more closely
follows individuals’ actions of updating their prior belief. For analyzing political decision–
making, the assumptions of Rational Choice Theory implies Bayesian behaviors of individuals
(Berger 1982). Frequentist and Bayesian methods typically produce the same results, but
Bayesian estimates tend to be more precise (Bayarri & Berger 2004, Lindley & Smith 1972).
An instance in which Bayesian statics are able to perform better is in overcoming dilemmas with Maximum Likelihood Estimation. Given that a uniform prior is used, the posterior
mode is equivalent to the Maximum Likelihood Estimate. If the distribution of the parameter
is multi–modal, then an MLE may become trapped in a local maximum and unable to find
the global maximum
15 .
In this instance, the likelihood of the parameter is improperly esti-
mated. Alternatively, there may be several maxima, which are not given through a single point
estimate. The Analysis of the full distribution circumvents this dilemma by fully expressing
the parameter space, allowing for analysis of various descriptive statistics of posterior estimates
instead of using point estimates.
15
Possible failures of an MLE are a result of the algorithm with which the MLE is approximated. There is not
a set of criteria for which convergence of the algorithm is guaranteed; the iterative process may diverge or settle
on a local that is not the global maximum. With the use of an Expectation Maximization algorithm (EM), the
estimate will converge, but can converge to minima, or a local maxima—convergence to the global maximum is
not assured, and the rate of convergence is slow (see Weihs, Mersmann & Ligges 2013)
42
Additionally, the use of probability densities in Bayesian analysis allows for easy transformation of results into prediction through the Predictive Probability Distribution:
Z
P (yn+1 , ϑ | D)P (ϑ | D)dϑ
P (yn+1 | D) =
θ
Given the data, a probability distribution of a new observation allows one to infer the likelihood
of future values. Combining the posterior distributions for the regression coefficients, the probability of a range of future observations can be expressed. As opposed to predictive inference
based on a 95% chance that a new observation will fall within plus or minus the margin of
error for a point estimate, the PPD gives insight into the relative degree of belief across possible
values of future data. The PPD also provides an opportunity for model checking. Using the
distributions of the betas, simulated response variables can be created under the assumptions of
the model. Comparing the simulated data to the observed gives insight to how well the model
describes the phenomenon of interest. For categorical data, the accuracy can be quantified
with the percent correctly classified: the proportion of observations that the model successfully
predicts (See Gill, 2013; Bland and Altman 1998; Jackman 2004 for a more rigorous treatment).
Bayesian analysis also offers a more intuitive means of making inferences from estimation: the Credible Interval. Discussing uncertainty in terms of probabilities, or degrees of belief,
offers a more natural interpretation (see Lecoutre et al. 2001). The lower and upper credible
bounds, (Cl , Cu ), are chosen such that:
Z
Cl
Z
P (β | y)dβ
0.025 =
and
Cu
P (β | y)dβ
0.975 =
−∞
−∞
The probability that β is less than Cl is 2.5%, while the probability that β is less than Cu is
97.5%. Together, there is a 95% chance that the parameter that generated the data is between
the two bounds. The Credible Interval expresses the most probable values through the Highest
Posterior Density region (HPD)16 :
16
An HPD region does not necessarily need to be continuous. In the case that the posterior is multi–modal,
the HPD may cover multiple disjoint intervals. However, a unimodal distribution is the most common, in which
case the HPD region is a single interval.
43
Z
Cu
P (β | y)dβ
Cl
The Bayesian Credible Interval says, given the data, there is a 95% probability that the parameter that generated the data lies within the HPD region. Uncertainty of the population
parameter is treated as a random variable, giving a range of plausible values as a probability
density, allowing for direct inference of the parameter value. A Confidence Interval at the α=.05
significance level states that:
P (θ̂1 < Θ < θ̂2 ) = .95
If repeated, the population parameter would be within 95% of the Confidence Intervals produced
by different data, with the distribution of the estimates approaching the true value asymptotically. The HPD regions allows inference about the range of parameter values that generated
the data, as opposed to a statement concerning the expectation of repeated analyses. The
Confidence and Credible Intervals differ in that the former concerns the probability of obtaining the observed data under the hypothesis, P (D | H), while the latter is the probability of
the hypothesis given the data, P (H | D). The distinction of the two is representative of the
different approaches, in which Bayesian inference concerns inverse probabilities, or inferring the
probability of an unknown cause given the observed effects.
4.4
Model Selection
Inherent uncertainty exists with model and variable selection: given k variables, there are 2k
possible specifications. Bayesian Model Averaging (BMA) allows one to address the uncertainty
involved in model designations and their respective parameters (For an overview of variable
selection problems and solutions, see George 2000, Bayarri & Berger 2004). A particular problem
with model selection is the possibility for coefficients to have different values or the opposite
relationship with the dependent variable, based on the different specifications. BMA allows for
the examination of the estimates and direction of relationships across models. The posterior
probability of a model can be calculated with Bayes Theorem: the probability of a model is
equal to the probability that the data was generated by the assumptions of a given model, and
then averaged over all possible models’ performances, a constant across specifications. That is,
44
BMA tests which Mγ is the most likely to describe the factors by which the data was created.
Given an uninformative prior and a sufficient sample size, the Likelihood Ration Test (LRT)
and BMA are effectively the same. The LRT compares the MLEs, which is the median of the
posterior; BMA defines the full probability distribution of a model, inclusive of the MLE.
The posterior model probability (PMP) or the probability of a given model conditioned
on the data, is calculated as follows:
P (D | Mγ )P (Mγ )
P (Mγ | D) = P2k
j=1 P (D | Mj )pr(Mj )
Z
P (D | Mγ ) = P (D | ϑk , Mγ )P (ϑk | Mγ )dϑ
(1)
(2)
θ
where
P (Mγ | D) ∝ P (y | Mγ , X)P (Mγ )
(3)
From (1), P (Mγ | D) is the PMP, P (y | Mγ , X) is the marginal likelihood of the model, and
P (Mγ ) is the model prior. The integrated likelihood (2) is the probability of the model across
all possible vectors of parameters, θk . If there are relativly few variables, the model space can
be enumerated; however, the calculation often has intractability issues or is too large to feasibly
calculate, in which case, the integrated likelihood is approximated using MCMC methods (see
Hoeting, Madigan, Raftery & Volinsky 1999, Wasserman 2000).
In addition to model specification, the marginal probability that a regressor should be
included can be calculated: Posterior Inclusion Probabilities (PIP). The PIP is the probability
that a given variable should be included in the model best able to describe the data (Tsangarides
2004). More concretely, a given variable vi ’s PIP, qi , is the sum of PMPs of all models that
include vi (Berger & Molina 2005, Scott, Berger et al. 2010):
qi = P (vi ∈ Mγ | D) =
X
P (Mγ | D)
(4)
j:vi ∈Mγ
PIPs can assist in determining which variables should be considered for model specification.
45
Although BMA is conditioned on the data and included variables, which can give rise to similar
specification issues, such as which factors to include in model averaging, the PMP and PIP are
able to inform the model selection process and assist with addressing uncertainty.
I used Bayesian Model Averaging in order to aid in specification of the mainstream
and vote switching models. There are a number of control variables whose results are mixed, as
well as some obscurity with the relative importance of left–right ideology versus policy positions
in voting behavior. The PIPs of the variables is given in 1. Since there are only 210 possible
specifications, the model space was enumerated. A Bayesian Adaptive Sampling algorithm was
used. The uninformative model prior was used, P (Mγ ) = BetaBinomial(α = 1, β = 1); the
prior for the regression coefficients is a Generalized g–prior designed for BMA of GLMs, with
α = .5, β = 7213—the number of complete observations—and s = 0 (For an overview of the
Bayesian Adaptive Sampling algorithm, see Clyde, Ghosh & Littman 2012).
External political efficacy and government satisfaction are included in both the mainstream and
Table 1: Posterior Inclusion Probabilities
Intercept
Efficacy
Gov’t Satisfaction
Age
Class
Education
Policy Preference
Income
Left–Right
Religion
Sex
Political Interest
Vote Switch
1.00 *
1.00 *
1.00 *
1.00 *
0.77 *
0.91 *
1.00 *
0.99 *
0.10
0.50
0.11
0.18
Vote Mainstream
1.00 *
0.94 *
1.00 *
1.00 *
0.65 *
1.00 *
0.36
1.00 *
1.00 *
1.00 *
0.16
0.11
* Variable included in the most probable model, given the data
switching models with the highest marginal probability, and is highly likely to be a component in
explaining the dependent variable. Age, education, class, and income were included in all three
of the models17 . Although religiosity has a high PIP, the variable was excluded from the analysis.
Given the lack of complete cases, and the decline of social cleavages and the influence of religion
in politics, the exclusion should not have a substantive effect. The results of the BMA concerning
the control variables aligns with previous research (Arzheimer 2009, Coffé 2005, Van der Brug
17
Model that including all variables with a PIP greater than .5 has the greatest predictive capabilities (Bayarri
& Berger 2004)
46
et al. 2000, Van der Brug & Fennema 2003, Van der Brug et al. 2005). One difference is
in the importance of sex. While the variable has been significant in some analysis, while
having no effect in others (Arzheimer & Carter 2006, Van der Brug et al. 2000, Van der Brug
& Fennema 2003). The PIPs of sex and political interest are relativly low, and therefore
are excluded from the final model specification. Interestingly, left–right is highly likely in
a mainstream versus outsider support model, while negligible for vote switching; and policy
preference is probable for vote switching and unlikely for mainstream models. The possible
different roles and explanatory capabilities of these two variables is discussed later.
For the regression, the model specifications are as follows:
P (yn ) = β0 + β1 Ef f icacy + β2 Government Satisf action + β3 Lef t − Right + β4 P olicy
β5 Income + β6 Class + β7 Education + β8 Age + Both vote switching and mainstream vote are binary variables, and therefore analyzed with
logistic regression. As a categorical response variable, party choice is modeled with multinomial logistic regression. For vote switching, policy preference is used and left–right ideology
is removed, the opposite is the case for mainstream voting, and party choice includes both
variables. To estimate the densities of the regression coefficients, MCMC methods are used.
The burnin period is 50,000 iterations, and then 300,000 iterations are run thereafter. In order
to reduce the autocorrelation of the chain, only every 20th iteration is stored, bringing the
number of effective iterations to 15,000. For a standard logistic regression prior (see Gelman
et al. 2014), all the independent variables are centered and scaled, such that the mean is zero
and the standard deviation is .51819 . A reference prior is used with a Cauchy distribution. The
location parameter is set to the MLE obtained from frequentist regression, and the scale is 2.5:
β~p ∼ Cauchy(β~M LE , 2.5), resulting in a prior density f (x) = (2.5π(1 +
x−βM LE −1
)) .
2.5
The prior
is weakly informative, using the MLEs as initial approximations, in order to increase MCMC
efficiency and assist convergence during the iterative process to construct the posterior; however, the distribution is quite diffuse, and, given the sample size, the information from prior will
have negligible effects on the posterior20 .
18
.5
The variables are recoded as follows: (xi − x̄) SD(x
i)
After rescaling, the variable ranges are as follows. Efficacy: -.87 to .62; Satisfaction: -1.16, 1.08; Policy: -.95
to 1.35; Left–right: -1.16, .99; Income: -.91, .84; Class: -.72, 1.11; Education: -.62, .75; Age: -.80, 98.
20
See Appendix E: MCMC Diagnostics for a discussion of robustness.
19
47
For the first and second hypothesis, it is expected that both efficacy and satisfaction
will be negatively associated with vote switching. The directionality of the policy variable
was not hypothesized, but it is expected that their will be a substantive relation. In the case
of mainstream voting, I expect that efficacy and satisfaction will have a positive correlation,
with the less alienated and more satisfied supporting the establishment. For party selection
within the outsider and mainstream categorization, I do not anticipate that efficacy will have
substantial effects on many parties. Likewise, satisfaction with the government’s performance is
not expected to have a large effect on party selection21 . Both policy preferences and left–right
ideology should have substantiative implications on party support: a negative association is
expected for left–leaning parties, where a positive relation should exist for conservative parties.
5
Results
5.1
Voters’ and Parties’ Ideology
Figure 3 displays overlap of voters’ own ideology and their perception of the party’s ideology
that they intend to support. Generally, there is a strong relation between one’s self placement
on a left–right scale and their selected party’s; however, the strength of the relationship varies
across parties. Of the mainstream parties, there is a fair amount of variability among the
correlations: PvdA was the third highest at .61, CDA the fourth at .55, while VVD was the
second lowest at .38. The overall highest correlation was with the Socialist Party at .67; the
lowest correlation of .35 is the SGP’s. Mainstream voters’ ideologies follow a similar pattern
to the correlations, where PvdA is the third most liberal, and CDA and VVD are the fourth
and second most conservative, respectively. The most liberal constituency belongs to the Green
Left, the most conservative to the SGP, and D66’s voters are the most centric of all the parties.
Overall, left–leaning parties are more closely aligned to their voters than the more conservative.
Notably, those that supported LPF see themselves as more centric than the upstart party: most
supporters place themselves as just–right–of–center with a six of 10, but mostly see LPF as a
nine out of 10.
When deciding which party to support, the voter–party ideological proximity also
21
Given the wording of the question, their could be a significant relationship with intention to support the
mainstream parties.
48
Figure 3: Parties’ and their Voters’ Left–Right Ideologies
10
0.30
10
0
10
0.30
0.00
2
4
6
8
Left−Right Placement
.43
10
0
10
0.30
0.20
0.00
0.10
Density
0.20
0.30
10
2
4
6
8
Left−Right Placement
.55
SGP
0.00
0.10
Density
0.20
0.30
Christen Unie
0.00
2
4
6
8
Left−Right Placement
.38
0.20
Density
0
VVD
0
10
0.10
0.30
0.20
0.00
2
4
6
8
Left−Right Placement
.54
2
4
6
8
Left−Right Placement
.61
CDA
0.10
Density
0.20
0.10
0.00
Density
2
4
6
8
Left−Right Placement
.67
Lijst Pim Fortuyn
0.30
D66
0
0.20
0.00
0
0.10
2
4
6
8
Left−Right Placement
.42
0.10
Density
0.20
0.00
0.10
Density
0.20
0.10
0.00
Density
0
Density
PvdA
0.30
Socialist Party
0.30
Groen Links
0
2
4
6
8
Left−Right Placement
.65
10
0
2
4
6
8
Left−Right Placement
.35
The solid color represents the parties’ supporters placement of the organization on a left–right ideological scale;
The lighter bars outlined in black are the party supporters’ self–placement on the same scale. The correlation of
the two is below. Parties are organized from left to right by their voters’ average left–right self–placement. Red
indicates outsider and blue mainstream.
49
10
appears in loyalty to one’s chosen party. Table 2 represents the flux between the party one
supported in the last national election and vote intention in the current election22 . Across all
studies, just under 65% of all respondents intending to vote indicated support a mainstream
party. CDA has the largest base of intended support, with PvdA a close second, and VVD
trailing with a little more than half as many supporters. A few short of the VVD, the fourth
most common vote intention is to not vote. Among outsider parties, the number of intended
votes is as follows: D66, other, GL, SP, CU, SGP, and LPF with the lowest numbers of vote
intentions23 . PvdA loses the most support to respondents not intending to vote in the current
cycle, followed by D66 and other. Overall, 23% do not plan to vote for PvdA again, and it
only gains 15% in new intended votes. Similarly, CDA and VVD lose 17% and 26% and gain
15% and 20%, respectively. The Greens gain from PvdA, but also lose to the Socialist Party,
who successfully pulls in new voters: 57% of SP’s intended support is from respondents not
previously supporting them. ChristenUnie picks up 12% of their total votes from CDA. SGP,
despite their low total support, is the largest beneficiary of new voters, gaining 28%, about half
of which are those who previously supported an “other” party; and the organization only lost
13% of their base. During LPF’s brief existence, it mainly drew from VVD, previous non–voters,
22
Although a voter’s intended vote may not manifest in actual support, it is an indication that one is flirting
with the idea of either staying the course or changing support
23
LPF is underrepresented as compared to the other parties, given that the party was only present in 2002,
then greatly diminished for 2003 and 2006
Table 2: Changing Party Loyalties: Past and Present
PvdA
CDA
VVD
D66
GL
SGP
CU
LPF
SP
Other
Not Vote
Total
PvdA
2829
92
27
101
10
1
1
2
9
68
194
3334
CDA
85
2983
192
44
6
9
19
2
4
25
136
3505
VVD
40
121
1464
68
2
0
0
9
1
16
99
1820
D66
148
51
51
492
3
0
3
0
1
28
73
850
GL
99
21
14
55
225
0
1
1
11
38
16
481
SGP
2
10
0
0
0
159
2
0
0
32
3
208
CU
2
38
6
1
1
7
244
0
1
7
11
318
LPF
20
9
61
3
7
0
2
1
6
2
31
142
SP
87
18
14
20
29
0
0
5
153
8
23
357
Other
113
65
52
34
3
2
8
24
7
216
69
593
Not Vote
251
219
95
57
9
6
5
12
9
42
1112
1817
The row is the party respondents supported in previous elections; the columns are the party they
intend to support in the upcoming election; the diagonal is numbers of those staying with their
previous party. The last row is the total number of those intending to support a party, and the last
column is the total of previous support.
50
Total
3676
3627
1976
875
295
184
285
56
202
482
1762
and then used–to–be PvdA supporters. In all, only 62% of those that did not vote in the last
election intend to do so again. Intended non–voters comprise 14% of the respondents across
studies. From the pool of previous non–voters that intend to re–engage for the current polls,
24% choose either PvdA, CDA, or VVD; the other 13% plans to vote for an outsider party.
When deciding to change party support, voters tend to move towards movements with a similar
ideological profile as their current party, i.e., a PvdA–D66 or CDA–CU switch is more common
than movement across the left–right spectrum. An individual’s level of efficacy, self–placement
on a left–right scale, and position in the policy space are relativly the same for individuals
switching within the mainstream, to an outsider party, within outsider movements, and moving
to the mainstream. However, on a zero to one scale, the average level of satisfaction for those
switching between and to a mainstream party is .35 and .38, respectively; one moving to an
outsider party has average satisfaction of .43, while switching between has an average level of
.45. As a one is the most dissatisfied, respondents that left the mainstream and shift between
non–establishment movements are more dissatisfied than their mainstream–supporting cohorts.
5.1.1
Vote Switching
Where table 2 gives some insight to voters’ movement, or lack
Figure 4: Vote Switching Regression
thereof, between parties, information on one’s previous vote and
current intention does not explain
the decision–making process behind
vote switching. Table 4 contains the
results from the regression of atti-
Intercept
Efficacy
Satisfaction
Policy
Income
Class
Education
Age
Mean
0.274
0.634
0.670
1.163
0.881
1.154
1.564
0.588
PCC:
Lower CI
0.257
0.559
0.596
1.024
0.776
0.992
1.350
0.516
.7833
Upper CI
0.292
0.722
0.753
1.320
1.002
1.338
1.801
0.676
Null:
Effective Size
9142.000
9072.000
8984.000
8857.000
9755.000
9183.000
9334.000
9542.000
.7826
tudinal, policy preferences, and demographics on the intention to support a different party24 . The analysis supports the first hypothesis, in
which efficacy is theorized the have
Figure 5: Expected Probability of Vote Switching
Min
-.5
0
.5
Max
Efficacy
0.290
0.256
0.215
0.179
0.171
Satisfaction
0.304
0.251
0.215
0.183
0.151
Policy
0.192
0.203
0.215
0.228
0.252
24
The posterior means are approximately equal to the MLE and are the most likely values of the regression
coefficients, while there is a 95% probability that the parameter is between the lower and upper credible bounds.
See Appendix E, Sensitivity Analysis for brief overview. See Appendix D, table 13 for the results of frequentist
regression and the MLEs
51
a negative relation with vote switching. There is a 95% probability that the odds of efficacy are between .559 and .722, with a mean
of .634. When one becomes increasingly inefficacious then the individual’s propensity to change
party support increases. Likewise, a more satisfied individuals is less likely to change their
party from the one which they supported last cycle: the mean odds are .670. Moving towards
a more conservative policy profile accompanies a growing probability of vote switching, with
mean odds 1.16. Although the posterior distribution of the policy preference coefficient is near
1, there is a greater than 95% probability that odds are greater than one. As for demographics,
older respondents are more likely to remain loyal to their party, while those who have attained
higher education switch with a greater prevalence. While the posterior distribution of both
class and income contain one, it is still highly likely the coefficient that generated the data is
positive for the former and negative for the latter. As for the effects across the variables, the
expected probabilities are located in table 525 . The most inefficacious have a 29%26 chance of
vote switching, while the most empowered are only expected to switch with a probability of
17%; the total change is 11.9. Satisfied respondents have a 15% chance of switching, where the
change for the most dissatisfied is double that of the most satisfied. An individual’s average
position in the policy space has less effect of vote switching, with a total change of 6% from
liberal to conservative.
The Percent correctly classified is 78.33%, which is approximately equal to the null
model’s PCC of .7826. Simulating a PPD from the regression, the probability density of a new
observation’s propensity to vote switch is approximately normal, with mean .217 and standard
deviation .007. The median of the simulated values for each individual under–predicts the
prevalence of vote switching—only ten were predicted to switch votes. However, when using
the median for the PCC, given the relativly low propensity to vote switch, it is unlikely that
the median of the replications for an individual will be 1, indicating a vote switch27 . Fully
expressing the distribution is more beneficial for evaluating the model. The percentage of false
positives is effectively zero; however, the rate of false negatives is 21.6%. The PPD for vote
switching can be found in figure 8. From the simulations, the density of the number of voters
25
Since all variables were re–scaled, the expected probabilities were recorded at the same values for the respective predictors: the minimum and maximum values for the variable; the mean, 0; and one standard deviation
above and below the mean, ±.5
26
The expected probabilities are generated using the posterior means of the respective coefficient.
27
See Appendix B: Vote Switching
52
who intend to vote differently from the previous election is plotted. The distribution is quite
accurate, with the median of the PPD is 1530, which is close to the 1538 actual vote switchers;
the standard deviation is 49. Given that total number of respondents is 6,845, the PPD is a
relativly precise. The probability that the model over–predicts the amount of vote switching is
51%. On average, the model correctly predicts the amount of vote switching, but there is still
a fair amount of uncertainty.
For the vote switching model, all the MCMC tests suggests that the chain converged
to the true distribution of the regression coefficients28 . The posterior means for the odds
of the coefficients are within .1 of the MLEs produced through frequentist regression. All
variables passed the Geweke and Heidelberger–Welch tests, suggesting that the chain reached
its stationary distribution. The autocorrelation of the chain is around .3 for all variables at a lag
of 20, which effectively decreases to zero before a lag of 100. The autocorrelation contributes to
slow mixing of the chain, which required a larger number of iterations and explains the decreased
effective sample size. The cross–correlations between the variables was low, with most below .1.
5.1.2
Mainstream Voting
As expected under the third hypothesis, political attitudes are influential components in the
decision to support the mainstream versus an outsider party29 . As an individual becomes
increasingly efficacious, the probability that they will intend to vote for PvdA, CDA, or VVD
is higher than for the inefficacious. The mean odds for the efficacy score are 1.33, with a lower
bound of 1.178 and an upper bound at 1.513. Similarly, as expected with the fourth hypothesis,
the dissatisfied are less likely to support the established parties. There is a 95% chance that
the regression coefficient’s odds for satisfaction are between 1.825 and 2.29, with mean 1.825.
While the vote switching model used policy preferences, the mainstream voting model uses
left–right positioning, which has posterior mean odds of 1.751. Although the control variables
where somewhat mixed for vote switching, they are more likely to influence mainstream voting.
The wealthy and more highly educated are more likely to support the mainstream, as are older
voters. Notably, while income has a positive relation, self–perception of class has a negative
relationship: an increase in net income is associated with supporting the mainstream, while
28
29
See Appendix E: MCMC Diagnostics for the results of tests.
See Appendix C for regressions by year
53
seeing oneself as upper class reduces the chance of voting for PvdA, CDA, or VVD.
The PCC for the mainFigure 6: Mainstream Voting Regression Table
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
Mean
3.031
1.337
2.049
1.751
1.373
0.843
0.455
1.806
PCC:
Lower CI
2.852
1.178
1.825
1.562
1.209
0.733
0.397
1.585
.733
Upper CI
3.230
1.513
2.290
1.961
1.551
0.977
0.528
2.053
Null:
Effective Size
8901.000
9085.000
8994.000
9201.000
9237.000
9109.000
9192.000
8893.000
.722
stream voting model is .733, which
is a marginal improvement over the
null model with a PCC of .722.
Comparing the simulated y values
with the observed, false negatives
are fairly rare, occurring 3.27% of
the time; the rate of false positives
is 23.4% 30 . The PPD is given in fig-
Figure 7: Expected Probability of Mainstream Vote
Min
-.5
0
.5
Max
Efficacy
0.701
0.724
0.752
0.778
0.784
Satisfaction
0.568
0.679
0.752
0.813
0.869
Left–Right
0.612
0.696
0.752
0.800
0.841
ure 8. The median is further from
the observed value than with vote
switching. The most–probable number of mainstream voters from the
PPD simulations was 5,438, as compared to actual number of 5,406;
however, the model still performs relativly well. The observed number is still well within the
HPD region, and given a sample of 7,485, a difference of 32 is relativly small. A higher rate
of false negatives is likely a consequence of the operationalization of left–right ideology. As a
general scale was used, and not an individual’s proximity to parties, there was a trade–off between accuracy and model parsimony31 . The standard deviation is 54, which is relativly precise.
Although a fairly sharp distribution, the model exhibits some bias, tending to over–predict the
number of mainstream voters. The probability that a value greater than the observed will occur
is 72%.
Table 7 is the expected probabilities of mainstream voting. Notably, satisfaction with
the performance of the government has quite a substantive impact on an individual’s propensity
to vote for PvdA, CDA, or VVD. The most satisfied are estimated to vote for a mainstream
party about 86% of the time versus the most dissatisfied individuals’ probability of 56.8%32 .
30
See Appendix C: Mainstream Voting for a discussion on incorrect classification
In future analysis, a more equal compromise between the two goals may be reached through a quadratic
in left–right self placement. The linear term would account for a general left–right self–placement, while the
quadratic term would account for relative extremity, as compared to the average.
32
The effects of satisfaction may be overstated. The question in the DPES regards approval of the government
31
54
As hypothesized, the efficacious will vote for a mainstream party more often. An empowered
respondent has a 78% probability of intending to vote for PvdA, CDA, or VVD, will the alienated
respondents’ probability is 70%. Liberal voters are anticipated to choose the mainstream 61%
of the time, and the most conservative support the establishment at an expected rate of 84%.
All variables passed the mean test of the Geweke diagnostic; and each regressors passed
the stationarity and half–width portions from the Heidelberger–Welch tests. Given the success
of the both, it is highly suggestive that the chain converged to the stationarity distribution for
the coefficient estimates. Similar to the vote switching regression, the posterior means of the
mainstream coefficients are nearly identical to the MLE estimate under frequentist regression.
A little more that five per cent of iterations were accepted, lower than the optimal rate. The
low acceptance rate could is likely due to slow mixing and an autocorrelation of .3 after 20 lags,
disappearing before 100. Cross–correlations follow a similar pattern as with vote switching,
with nearly all below .1.
5.1.3
Party Choice
Overall, the effects of efficacy and satisfaction on party choice are less pronounced than with
vote switching and mainstream support; in deciding which party to support within the mainstream or outsider organizations, ideology is the most prominent component. Unexpectedly,
an individual’s political efficacy is significant for supporting PvdA and CDA, while satisfaction
is highly likely to relate to all three mainstream parties. Efficacious and satisfied individuals
are more likely to move from “undecided” to selecting one of the mainstream parties33 . However, the distinction of which to vote for comes with ideology. Support for the PvdA is highest
with liberal respondents; choosing the VVD or CDA is associated with a more conservative
self–placement. In agreement with previous work on party choice, left–right ideology is a large
component in deciding between alternatives. Similar to left–right placement, the PvdA gains
those supporting traditionally liberal policy solutions, with CDA and VVD winning those with
conservative positions. PvdA has greater support from the lower classes, while VVD and CDA
since the last election. Some outsider parties have gained access to coalitions in various elections, but at least one
of the major three parties is represented in all the governments since 1971. Given the phrasing of the question, a
portion of the overall effect is likely due to a reaction specific to PvdA, CDA, or VVD, depending on which was
in power at the time of the survey.
33
The relevance of efficacy for mainstream party support likely originates in that the efficacious are more likely
to have decided what party they will support.
55
benefit from the upper classes, and income follows the same pattern as class. Older voters are
more likely to support all three, although more strongly related to CDA.
The choice of which outsider party to support is influenced by similar factors as selecting between mainstream alternatives; policy preferences and left–right ideology are large factors
in the party choice decision. External efficacy is related with supporting the Socialist Party,
CU, LPF, and other—the influence of efficacy on party choice is more than hypothesized. The
efficacious are more likely to support CU, and less likely to support the others. Satisfaction has
a positive relation with D66, and a negative relation with CU, LPF, and other. Placing oneself
further left on the ideological spectrum is a significant factor behind supporting D66, GL, SP,
and other; conservative voters have a greater tendency to support CU, SGP, and LPF. However,
Figure 8: Posterior Predict Distributions
0.004
0.000
Density
0.008
Predicted Distribution of Vote Switching
1300
1400
1500
1600
1700
Numer of Vote Switchers
0.004
0.000
Density
0.008
Predicted Distribution of Mainstream Voting
5200
5300
5400
5500
Numer of Mainstream Votes
56
5600
it appears that separation occurred between SGP and left–right placement, as well as policy
preferences. Given the unique ideology and the party’s placement as the most conservative, it
is to be expected that ideological variance within its voters is relativly low; the positioning of
the party in the policy and ideology space make it such that their constituency is highly homogeneous. The relationship between party choice and policy preference is the same as ideology.
Higher classes favor the D66 and GL, while the lower are more partial to the Socialists and
LPF. Income is positively related with support for D66, and negative for all others except the
SGP. Individuals with higher education tend to support the GL, SP, CU, LPF, with the less
educated leaning more towards other. One’s age is mostly unrelated, except for LPF and SP,
which wins more voters from older age categories.
6
Discussion
Following the trends across parliamentary democracies, the Netherlands has faced increasing
electoral volatility and decreasing support for mainstream parties. Shifting dynamics raise questions about the origins behind the decline of traditional parties and fading loyalties. Answers
to the causes of these changes have focused on ideological components; this paper proposes
a social movements approach, in which the shifts are examined as unconventional behaviors
whose cause is rooted in attitudes. Value orientations have seen a shift, whether through post
materialism or New Politics, but there may be another factor underlying the changes: individuals’ evaluation of their influence within, and satisfaction towards government. As with work
on participation, the inclusion of expressive motivations through political attitudes can further
the understanding of party choice. There is more to a voter’s evaluation of parties than their
policy stance and general ideology with respect to their own; one’s satisfaction with parties
and sentiments of political power enter into the calculation. As an individual is motivated
to participate based on the belief that their action will have influence, they seek to maintain
that influence within a political organization. Where dissatisfaction and alienation motivates
unconventional participation outside of the system, it influences non–traditional choices within
the electoral system. Political powerlessness and disillusionment contribute to the changing
status quo. Non–institutional action and participation within a political system share common
influences, in which social movements and electoral behavior are rooted in similar motivations
57
Party Choice Regression Tables
Mainstream
Mean
Party:
Intercept
Efficacy
Satisfaction
Left–Right
Preferences
Income
Class
Education
Age
1.2
1.291
1.295
0.128
0.404
0.942
0.565
0.452
1.96
.025
PvdA
1.095
1.077
1.1
0.103
0.333
0.792
0.463
0.371
1.638
.975
Mean
1.313
1.528
1.535
0.158
0.492
1.123
0.692
0.554
2.351
1.165
1.425
3.155
4.293
2.063
1.105
1.125
0.545
3.028
.025
CDA
1.059
1.198
2.65
3.477
1.692
0.923
0.924
0.445
2.541
.975
Mean
1.276
1.695
3.789
5.288
2.507
1.31
1.385
0.662
3.62
0.487
1.191
1.702
9.941
1.613
2.17
2.996
0.878
1.393
.025
VVD
0.43
0.985
1.405
7.765
1.297
1.778
2.368
0.704
1.14
.975
0.552
1.442
2.064
12.71
2.015
2.62
3.808
1.092
1.694
Outsider
Mean
Party
Intercept
Efficacy
Satisfaction
Left–Right
Preferences
Income
Class
Education
Age
Party
Intercept
Efficacy
Satisfaction
Left–Right
Preferences
Income
Class
Education
Age
.975
Mean
0.391
1.083
1.448
0.606
0.339
1.307
1.792
0.791
0.787
.025
D66
0.342
0.856
1.152
0.461
0.254
1.045
1.368
0.609
0.611
.975
Mean
0.108
1.154
1.176
0.05
0.519
0.733
1.694
1.923
1.276
.025
GL
0.086
0.862
0.9
0.033
0.368
0.574
1.235
1.379
0.956
0.446
1.356
1.84
0.813
0.451
1.624
2.321
1.047
1.004
0.006
1.522
0.884
51.15
31.84
1.308
0.482
1.136
0.849
SGP
0.003
0.933
0.527
23.289
15.855
0.738
0.27
0.632
0.516
0.011
2.519
1.474
114.688
62.563
2.272
0.858
2.067
1.39
.975
Mean
0.086
0.581
0.767
0.049
1.926
0.649
0.574
3.351
2.15
.025
SP
0.067
0.431
0.57
0.032
1.315
0.482
0.406
2.337
1.544
0.135
1.52
1.541
0.073
0.75
0.949
2.342
2.759
1.732
0.109
0.791
1.028
0.076
2.851
0.871
0.826
5.037
2.983
0.08
1.935
0.706
4.84
10.15
0.788
1.031
1.578
1.277
0.031
0.384
0.363
1.789
8.45
0.575
1.528
2.687
1.939
LPF
0.021
0.238
0.231
1.02
4.719
0.356
0.888
1.543
1.246
0.044
0.61
0.56
3.17
16.156
0.902
2.716
4.725
3.113
0.228
0.654
0.446
0.497
0.678
0.633
0.968
0.543
1.015
Other
0.195
0.492
0.343
0.351
0.485
0.482
0.695
0.394
0.757
0.266
0.87
0.595
0.691
0.941
0.831
1.313
0.762
1.367
The reference category is“undecided.”
58
.025
CU
0.062
1.387
0.504
3.154
6.55
0.553
0.698
1.05
0.902
.975
0.103
2.726
0.991
7.407
15.905
1.099
1.537
2.342
1.764
and have the same goals, forming a repertoire of unconventional behaviors. The analysis reaffirms the significance of ideology and policy preferences in vote choice, while supporting a social
movements approach. Vote switching is a reaction to disaffection and alienation within one’s
party, while outsider party support stems from general disillusionment and powerlessness within
government more broadly; changing party support is a movement against a singular party, while
leaving establishment parties is a movement against the mainstream and the political status
quo.
6.1
Vote Switching
The values shift that has occurred across Europe does not necessarily lead to a shift in political
alignment. Given the influence of ideology, the question arises “when will an individual decide
to change their party support?” One may vote for another party once their desired policy
position is far enough away from their current party’s. However, this raises additional questions
about the factors that influence an individual’s tolerance for a disconnect between themselves
and their party. What magnitude of difference would be required for an individual to change
loyalty? This project has found support for a different answer: the decision to change party
support is motivated by more than policy preferences, with an individual’s perceived political
power and affect impacting the choice.
6.1.1
Vote Switching and Individual Influence
There is support for the first hypothesis: when deciding to change party support, individuals
evaluate their political power. Parliamentary democracy grants voters a greater range of political options, especially in the Netherlands; however, not all options are evaluated equally.
After having decided to support a new party, there are clear alternatives. The choice of which
new party to support is based in instramental and expressive concerns, as well as the trade–off
between them. The relative importance the voter places on these factors influences to which
party one will switch. Given a range of possibilities, the voter may behave more sincerely, selecting a party that is more ideologically aligned with themselves, or they may support a party
whose organizational structure is more favorable than their current party’s. Although the vote
switcher may base their new party choice on spatial considerations, the motivation to switch is
59
rooted in their inefficacy.
Pursuing a greater ideological agreement, the voter may engage in vote signaling with
the intent of influencing their most–preferred party, or one may be less willing to compromise,
selecting a more ideologically pure organization. Signal voting can explain the relevance of
policy preference, but not left–right ideology, for vote switching, as seen in the model averaging.
The use of hard issue voting for changing one’s party supports previous work on the relevance
of policy preferences on party defection, following the expected results of expressive voting
behaviors (Erlingsson & Persson 2011, Franklin et al. 1994). There is a dependency between
efficacy and policy preferences, with both factors influencing the decision–making process behind
changing loyalties. Significance of both alienation and specific issue–area concerns suggests that
there is a trade–off between instrumental and expressive motivations. In supporting a political
party, one reduces their direct influence in order for the aggregated support of all members to
affect policy; the individual is unable to single–handedly control the direction of the party, in
return receiving beneficial—although not necessarily the most preferred—policy outputs. One
may stay with a party that has alienated them, but is large enough to influence policy, or one
can switch to a party more aligned with their positions, but unable to enact legislation. The
question becomes when an individual decides that their influence within the aggregate is too
little, and begins looking at other parties; what determines one’s willingness to trade expression
for outputs?
If one desires to send a specific message of support for a policy position, then they
may abandon their previous party to register their support for another party’s stance. When
engaging in signal voting, one will use hard issue voting, increasing the significance of policy
preference. In the case of signaling, the voter is likely more instrumentally–focused, in which the
goal is influencing policy outputs: the individual is attempting to alter the position of a party
and affect future legislation. When seeking greater ideological similarity, the voter is expected to
be expressively motivated, with a greater desire for a more purely representative organization. In
such a case, inefficaciousness contributes towards an emphasis on expressive considerations, with
the extent of one’s alienation influencing the amount of policy distance one will tolerate between
themselves and their chosen party. While similar behaviors, signaling and switching would
appear to have separate motivations—one instramental and the other expressive—but share a
similar goal of maximizing their efficacy. A desire to restore influence adjusts electoral behavior
60
in the same way that it promotes unconventional extra–system participation. The restoration
of influence can take the form of selecting a party that has the closest ideological proximity, or
in successfully signaling one’s former party. Alternatively, one may seek an organization that
grants more control to individuals.
An individual can also be motivated to switch based on party structure and the relation
between party elites and the membership. Not necessarily in support of a specific ideology or
policy, the voter is more concerned with restoring their ideal level of citizen control. Alienated
by a party they feel has taken too much power from its supporters, one searches for a party that
will uphold its constituency’s desires. These sentiments may manifest in perceptions that the
party elite have too great control within the organization, and therefore the voter searches for
more grass–roots parties and political outsiders. When individuals maintain a sense that they
have control over a political party, there is no need to change allegiance. There may be new
policy areas emerging, or old ones changing, but the perception that one has a say in government
affairs reduces the desires to change support. The efficacious perceive that they are able to shift
the party’s stance in their desired direction. However, when a sense that the individual has
lost influence in the organization begins to grow, they are more likely to search for alternatives.
Once one no longer has faith in party leaders, and no means within the organization exist for
reform, a member leaves the party; the voter is less concerned with the party’s ideology, as they
see the leadership is corrupt or untrustworthy (Levin and Eden 1962). Instead, the individual
will not support the current parties—or party—in government, but will cast their ballot for
the out–of–power mainstream party. Although inefficaciousness alone may not be enough. An
alienated individual may not change support, given that their policy preferences still align with
the organization’s. A decision to vote switch is an evaluation of one’s political power, their
ability to influence government, and their desired policy outcomes.
6.1.2
Vote Switching and Satisfaction
As with the first, there is support for the second hypothesis: it is more probable for a dissatisfied
vote to change loyalty than a satisfied one. Voters may also be motivated to change allegiance
due to their dissatisfaction or animosity towards their current party. Out of efficacy, satisfaction,
and policy preferences, government satisfaction had the largest effect on the expected probability
61
of vote switching. While the implications of political powerlessness may in part be conditioned
on policy differences, it is possible the dissatisfaction is independent of ideology. As with
motivations concerning one’s political power, affective considerations can also be seen as a
trade–off between instramental and expressive motivations. Even if the voter’s preferences
agree with the organization, if there is a sufficient level dissatisfaction, the individual may
change their support regardless of the ideological concordance. A voter may use their ballot
to register discontent with the party, but in doing so, reduces the organization’s ability to
create policy. This trade–off can be seen as a decision between registering support for an
organization and the desire to limit a party’s electoral mandate, as suggested in formal modeling
(Myatt 2015). How one reaches their decision depends on their relative level of disillusionment
with the organization. If more instrumentally–focused, one will not switch from the party with
which they are dissatisfied. However, when a voter is less concerned with casting a pivotal
ballot, and their disaffection increases, then they are more likely to cast a vote in order to
register their discontent. Where the most satisfied are only expected to switch 15% of the time,
those most dissatisfied have a 30% chance of switching.
The process of deciding on a new party can be seen through the case studies of the CDA
and VVD defections. One that feels they are powerless within the CDA has two alternatives:
the CU and SGP. If one desires to send a signal to the CDA, then support for the CU becomes
more attractive, while transferring to the SGP seeks greater ideological purity. Switching to
the CU is more instrumentally motivated, while the SGP is expressive. One may desire to leave
the mainstream party, but still concerned with ideological proximity, the voter can use the CU,
which, from the party choice regression, gains support from disaffected voters. Interestingly,
CU voters are more likely to be efficacious, raising the possibility that supporting the party can
restore a sense of political power. Alternatively, individuals who feel their vote more effective
may expect that a transition to CU from CDA will be met with the CDA reforming the win
back voters—the Christen Unie voter perceives that their signal will have the desired result and
alter the CDA’s future behavior. As the Christen Unie has served in governing coalitions, it
can serve as a balance between the expressive desires for a closer ideological proximity, while
maintaining the possibility that the organization will be able to enact policy. Alternatively,
one whose expressive motivations outweigh instramental may choose the SGP. Given that the
Reformed Political Party, currently the oldest party in the Netherlands, has never been in a
62
coalition, the likelihood that supporting the SGP will yield policy results is nearly zero. An
SGP voter is less willing to accept ideological distance in return for policy, instead electing
to support a more ideologically pure party that will be unable to enact their policies. A vote
for the SGP, as opposed to the CU, is more purely expressive act, registering support for the
policies of the party, less concerned with the ability to implement the policies. Such a trade–
off provides an explanation as to when Swedish voters will leave a traditional organization in
favor of the Pirate Party. While not abnormally dissatisfied (Erlingsson & Persson 2011), the
inability to influence the mainstream’s position on intellectual property pushes one towards an
unconventional party. For the case of the VVD, the mainstream party lost numerous voters to
LPF, while it was active. Lijst Pim Fortuyn offered a stronger position on immigration than
the VVD, creating the possibility of signaling to VVD to change its stance. However, LPF
may have also benefited from those that felt powerless within the organization and dissatisfied
with the VVD; the Pim Fortuyn List benefited from both the alienated and the disillusioned.
Therefore, LPF can serve both a signaling function, as well as source to register discontent.
When deciding whether to switch to a new party, there is more to the choice then ideology: an individual’s efficaciousness and level of satisfaction both influence the decision–making
process. Vote switching is not a non–ideological process, but neither is it solely ideology–based;
voters do not change party support in order to maximize their ideological similarities, instead
motivated by their inefficacy or satisfaction. The analysis supports previous works on the importance of policy for party support, but suggests that it may not be a complete picture. A voter’s
decision to vote switch depends on the importance they place on expressive and instramental
motivations. One may behave strategically or cast a sincere vote, depending on their satisfaction with a given party and their belief of the political power. When inefficacious, one seeks to
maximize their influence with the government. This desire can manifest in voting sincerely and
supporting the party closest to themselves ideologically, or in voting for an organization that
most upholds the individuals democratic norms or representative government. Inefficacy can
result in more sincere and less strategic voting, but the motivating factor is not one’s ideology;
the decision to vote switch is based in alienation and disaffection, which can lead to either more
spatially–focused behaviors or the desire to register discontent.
63
6.2
Mainstream Voting
While elections in advanced democracies have become increasingly volatile, decreasing voter
loyalty is not the only shift that has occurred. The dominance of the mainstream, established
political parties has a variety of new challengers; traditional political parties’ preeminence has
been diminished within the Netherlands, as well as across Europe. Decreases in established
parties’ support has been explained through values shifts and re–orientation of ideology, in
which there is a growing demand for parties other than the mainstream organizations, with
growing support for anti–immigrant parties of particular interest. However, this paper proposes
that the role for ideology may not be as large as expected, supplanted by political attitudes.
Ideology is a factor in mainstream versus outsider voting, but the decline of established parties
may be an expression of the inefficacious electorate, whose main concern is not sincere, spatial
voting. While new movements’ left–right positioning is heterogeneous and their politics cannot
be classified within single family, these parties’ support may have origins in the same factors.
The growth of outsider political parties can be seen as a social movement, whose origin lies in
alienation from and disaffection towards the mainstream.
The mainstream versus outsider party decision can be an instance in which ideology is
not the main consideration34 . Both power dynamics and voters’ sentiments towards government
have repercussions for party–type selection. From the BMA, positioning in the policy space has
less importance for outsider voting, while general left–right ideology is more important. This
could be due to party–type selection being based on a more heuristic evaluation of parties.
Instead of engaging in hard issue voting to select a party, one uses the less intensive soft issue
voting. A voter seeks an exit option with which they generally align, but specific policy stances
are less important. While one may behave spatially after deciding to vote outside the mainstream, the choice of which party type to support is not rooted in ideology. Where alienation
and disillusionment motivated unconventional behaviors outside of the political system, so will
they influence non–traditional voting.
34
For a comparison of an ideology–only and attitude–ideology model of mainstream voting, see Appendix C:
Mainstream Voting With and Without Attitudes
64
6.2.1
Efficacy and Voter–Party Power Dynamics
Supporting hypothesis three, inefficacious voters tend to support outsider parties more so than
empowered voters. Despite their sense of political powerlessness, many alienated voters turnout
to the polls; however, when casting their ballot, inefficacy influences their decision–making.
Voters who feel that they have little influence within government seek alternatives to the established PvdA, CDA, and VVD. Moving from the most efficacious to the least, the propensity to
support one of the mainstream parties decreases eight per cent, falling from 78.4% to 70.1%.
When one feels that the mainstream does not respond to their desires, that they do not have
influence within the organizations, they begin to look elsewhere for representation. When alienated from established politics, one becomes more likely to support a different type of political
party, one that is closer to its constituency and promotes the power of its supporters. The
inefficacious segment of the electorate does not behave in a traditional ideologically–based understanding; instead, they support movements that will re–empower them within the political
process. Underlying the transition to unconventional voting behaviors is a sense of political
powerlessness, that one does not have influence within their respective party or the government
as a whole. Differing from eroding party–specific loyalties, anti–establishment behaviors of the
alienated target status quo politic and the mainstream more broadly, with the goal of restoring
democratic norms of citizen control and general representativeness. In order to capture a plurality of voters and create governing coalitions, traditional catch–all parties require a big–tent
nature. While growing the base of supporters increases an organization’s political power, an
individual voter may sense that the organization has become too big for their voice to have any
meaningful impact. In reaction to the marginalized role of the individual, the alienated voter
searches to restore norms of representative democracy outside of the establishment. Responding
to cartelization or a perception of an oligarchic elite of the mainstream parties, a voter uses
unconventional behavior, acting in a social movement–like context.
Use of a social movements approach may be able to explain more extremist, and
possible anti–democratic politics. Voters’ reactions to their power relationship with parties and
perceptions of cartelization contributes to the role of relative ideological extremeness in choosing
outsider parties. A farther left or right voter will not feel represented by the centrist mainstream;
the organization is not one of like–minded voters, and may shun members that are relativly
more extreme. The diminished influence within traditional parties and possible ostracization
65
may motivate one to support an outsider political party. Combining left–right proximity with
institutional alienation can give insight into the transition from traditional politics to more
extreme outsider parties. Given that an individual has been estranged form traditional politics,
they can be pushed further towards the radical right or left. While outside the scope of this
project, the social movements approaching to voting behaviors can also lend an explanation to
radical politics. In the non–institutional component of the unconventional behaviors repertoire,
alienation contributes to political violence, such as riot participation (Paige 1971); but is there an
institutional counterpart to violence, and possible revolutionary behaviors? Polarizing parties,
those that are not anti–democratic, but more antagonistic (Capoccia 2002) may become more
appealing to the individual. If the sense of powerlessness is strong enough, on may turn to more
anti–system organizations. As efficacy can be present in authoritarian government through
symbolic gestures of citizen control (Coleman & Davis 1976), an individual facing extreme
alienation may turn to anti–democratic parties to restore their influence.
6.2.2
Satisfaction and Affective Motivations
There is also support for hypothesis four: the more dissatisfied a voter, the more likely they are
to support an outsider political party. The results supports previous work on dissatisfaction and
voting. When individuals are dissatisfied with mainstream movements, they change their party
loyalty or abstain in future elections (Miller 1974). As a voter becomes increasingly disaffected
with the government, support for the PvdA, CDA, or VVD decreases. An individual’s negative
sentiments towards the current government can stem from dislike of the policies that they have
enacted, or the way in which they have conducted themselves. As with Labzina and Schoflield
(2015), disapproval and lack of trust grows support for non–traditional parties. When no
alternatives exist, then the voter has the option to no longer participate in the electoral process,
or if their disaffection is not great enough for them to withdrawal from politics, the individual
will continue to support the organization with which they are unsatisfied. However, when an
exit option is available, then the prospective voter will change their support to the alternate,
non–establishment party. A more general dislike of the government—not necessarily specific to
a certain policy or action—has the possibility of altering how an individual evaluates the exit
option, or how they select among a plurality of options. Disaffection with mainstream politics
reduces the emphasis one places on spatial factors. Policy preferences are less important for the
66
decision to support an untraditional party, and the effects of left–right position are less than
that of satisfaction.
When dissatisfied with the mainstream, ones affect can become more important in
their decision making, in that the individual uses an exit option with which there may not
be a high degree of ideological similarity, but they have more positive sentiments towards the
organization. When negative affect towards government does not arise from a single cause, or
when there is not an especially salient origin of the disaffectedness, then a particular restorative
action—and therefore a specific exit option—is unclear; an obscured negative attitudes leads to
an obscured course of action, in which the party best able to remedy one’s dissatisfaction may
be unknown. In such a case, an individual will select an outsider party with less intent, using
softer decision–making. Their action is meant to register their discontent, but not necessarily a
solution—they engage in a sort of anti voting, in which there is a more general anything–but–
the–mainstream sentiment. The effects of government satisfaction on mainstream voting may be
a result of this anti sentiment, in which the voter is responding to cartelization or another source
of frustration. Rejection–based political action, motivated by general dissatisfaction, supports
the tendency of voters using less intensive evaluations engaging in more symbolic behaviors
(Pollock III 1983), like that seen in the 2009 Greek elections, in which the Greens were used
as a negative vote (Vasilopoulos & Demertzis 2013). Such actions would not have a clear goal
other than a kind of punitive or negation–orientated action against the party that is the origin
of the voter’s low satisfaction, supporting Thomposon and Horton’s (1960) work on political
alienation.
6.2.3
Left and Right–Leaning Voters’ Relation with their Parties
While outsider voters tend to be extreme ideologically—as to be expected—conservative voters
are less likely to support an outsider party. Two of the mainstream parties, the CDA and VVD,
being more conservative, while the outsider parties are split between left and right–leaning, the
D66, Socialists, and the Greens have more supports than the more conservative organizations:
LPF, Christen Unie, and the SGP. The relationship between further–right placement on the
ideology scale and support for mainstream parties may be a result of these two factors; however,
there is a possible interesting implication of the relationship between more left–leaning voters
67
and their parties. As for the higher support of mainstream parties among conservative voters,
right–leaning individuals are not particularly more or less efficacious than liberal voters: there is
effectively no correlation between efficacy and left–right self placement. Therefore, the difference
in party type support is not likely a factor of different levels of alienation between the two
ideological groups. However, there is a difference between the satisfaction levels of liberal and
conservative voters35 . When dichotomizing the left–right scale into left and right–leaning, as
well as satisfaction and efficacy into above and below average levels, there is a difference in
the level of satisfaction between the two ideological groups of voters. Given the data from the
DPES, individuals more liberal than the average are almost surely more dissatisfied than above–
average conservative voters. This difference suggests that the relationship between voters and
their party—and the mainstream more broadly—is influenced by more than ideology; and when
deciding whether to support traditional, established parties, the more disaffected liberal voters
may be more partial to outsider movements. A more dissatisfied liberal base of voters may be
able to explain, in part, the mechanism for the greater numbers of liberal outsider support.
Out of the 5,446 individuals that intended to vote for a left–leaning party, 33.4% intended to
support an outsider movement; among the total 6,417 intended conservative party voters, only
11.2% indicated they would select an outsider party.
It is possible that liberal voters are more attracted to non–mainstream movements,
or the relation is a result of the ideological composition of the mainstream; however, the more
disaffected liberal constituency can have consequences for how left–leaning parties and voters
relate within one another. Although not particularly more or less likely to be alienated, the left–
leaning voter tends to have lower affect towards the government then the conservative voter36 .
The disconnect may arise from the relative salience of traditionally left and right policies, and
how to electorate perceives the extent to which parties have acted on these issues. One could
feel that environmental regulations have not been adequately addressed by the PvdA—the
mainstream left party has not fulfilled the voter’s policy desire for environmental policy output.
For the voter, there is a clear exit option: the Greens. Similarly, for issues regarding income
inequality or the Dutch parliamentary democracy, the Socialist Party and D66, respectively,
serve as alternatives. Conservative voters may feel that the CDA and VVD are delivering
on their most salient political issues, and therefore are more satisfied and do not feel a need
35
36
See Appendix C: Mainstream Voting
See Appendix C
68
to support an outsider party to the same extent. Meanwhile, the PvdA has not sufficiently
implemented policy on its supports’ more important issue areas, so their voters become more
dissatisfied and begin looking at the possible alternatives with greater frequency. The effect of
salience can be interpreted to have the opposite cause: instead of supporting outsider parties
because the mainstream has not addressed important issues, in supporting Arzheimer (2009),
the increased salience of policy issues drives voters towards a more ideologically pure outsider.
The dissatisfaction may arise from the PvdA’s inability to create legislation, a perception of
cartelization, or that the PvdA has moved towards the center in an effort to grow its vote share
and gain supporters from the other two mainstream parties. The role of ideology in choosing a
traditional versus non–traditional party could be a function of the extent to which liberal and
conservative voters feel the mainstream has address their respective issues of focus.
While both efficacy and satisfaction with government can provide motivations for outsider support, they may operate as two independent factors, with each supplying a different
behavior from the unconventional repertoire. A sense of political powerlessness and disaffection
can create desires to restore influence or incumbent affect, respectively. However, alienation
from politics may promote a more strategic behavior, in which the individual engages in more
calculated political actions. In order to force a party to address a policy or change its position,
the individual must select which party they will use to cast a signal vote. The decision to send
a more specific message with a ballot requires more attention to be paid to what signal to send,
and how the individual perceives the party will respond; one must communicate their desire
effectively. If attempting to adjust a party’s stance, then one would need to select the correct
party with which to send the message. Additionally, the individual may base their action on
how they believe other voters will act. In the case of regaining influence, the voter would need
to select a party emphasizing representation and a stronger role for party supporter, and then
determine if their most–preferred party would properly receive the message that they need to
re–connect with their base.
Alternatively, an affective motivation may not promote strategic behaviors to the same
level as inefficacy. Instead of a more costly, calculated messaging, disillusionment with the
mainstream could promote more of an anti–vote: a ballot cast without any particular intended
outcome, but more of an act of registering discontent. The individual will still heuristically
choose an exit option that generally aligns with their ideology, but they will not seek to promote
69
a specific policy element of the party. One casting an anti vote would be showing their support
for untraditional, anti–establishment and mainstream politics. Such a voter would be inclined to
support a political outsider to change the status quo; the voter registers their support for their
type of politics. One may have more punitive motivations, in which high levels of dissatisfaction
and negative sentiments towards the mainstream leads the individual to vote for a party that
will punish the establishment, or otherwise challenge the current political climate. Dissatisfied
with traditional parties, the voter may have more positive feelings towards an outsider party
and its politicians, in which case, they would like to see the more favorable candidates succeed.
Given the relativly large effect of government satisfaction on mainstream support, the Dutch
electorate may be more affectively motivated, expressing a general disaffection as opposed to
signaling.
Growth of new politics and declining mainstream support can be seen through a social movements perspective, in which untraditional voting behaviors are similarly motivated
as untraditional extra–system action. While ideology is not irrelevant, it is not the motiving
force behind the changing electoral dynamics. Voters’ alienation and disillusionment provide
a source for the decision to support outsider or anti–establishment politics. While not tested
in this paper, there is a potential that a social movements understanding can also contribute
to work on political extremism. Likewise, interactions between efficacy and satisfaction can
further the development of the institutional arm of the unconventional behaviors repertoire.
6.3
Party Choice
Whereas efficacy and satisfaction are influential in the decision–making process behind vote
switching and outsider support, to what extent are these factors incorporated into specific party
choice decisions? The role of non–ideological expressive or attitudinal is more ambiguous when
choosing between alternatives. It was expected that after one had decided to cast a ballot for a
mainstream or outsider party, the significance of political attitudes would give way to ideological
and policy–specific factors. While that is the case for most parties, the extent to which efficacy
and satisfaction remain relevant for specific movement selection was unanticipated. Within the
mainstream, efficacy is quite probably related with a vote for the PvdA, CDA, and there is
a fairly large probability of a relation with VVD; there is a very probable relation between
70
satisfaction and intention to vote for all three of the parties. Given the relevance of these
two factors for mainstream party type selection, it would follow that they are also related with
choosing an individual party, as well. For the outsider parties, the attitudinal variables have less
influence. Efficacy is only influential with three of the seven possibilities, while satisfaction has
a high probability of being related with four. Interestingly, the Christen Unie is both positively
associates with efficacy, while negatively associated with satisfaction. This may be in part due
to CU serving as a clear alternative for CDA voters. When disaffected with the CDA, it appears
that voters may turn to the CU as a signal, sensing that their support for the Christen Unie
may be able to create the desired outcome.
Generally, the role of efficacy was expected to be minimal, but I did anticipate that
one’s sense of government responsiveness would be important for D66, the party advocating
radical democracy and more direct influence for citizens. It would seem that such a platform
would speak to the alienated, but that is not case. In fact, the probability that efficacy is
unrelated to D66 support is the highest among all the outsider parties. While founded as a
movement for greater democratization, it is possible that voters’ evaluation of the party has
shifted. The D66 has the greatest vote share of all outsider parties, has been in several governing
coalitions, and has moderated their platform since 1966. These components may have altered the
D66’s status with voters, possibly being seen as another established, traditional party. Softening
of the radical democracy platform, taking a more catch–all approach, may have decreased the
party’s credibility with the alienated electorate. Support for D66 is unlikely to have a relation
to one’s perceived political power, and very well may have been incorrectly classified as an
outsider party and should be considered a part of the mainstream in future analyses.
6.4
Electoral Earthquakes
When an exit option appears for the first time in an election, there is the potential to have a
large electoral shift. A large number of voters whose levels of dissatisfaction and alienation are
high creates an optimal environment for a change in party dynamics and a high level of volatility.
The disaffected or politically powerless have stayed with their party due to no available exits,
due to a sense of civic duty, or that their currently party is the best of the available alienates;
other voters may have dropped out of the electoral process, given increasing frustration and
71
perceptions that their actions cause no reaction. But, when an exit option becomes available,
the disaffected and or alienated are given another option, and the previously withdrawn can be
motivated to return to politics. There is a possibility that large numbers of voters will take the
exit option, abandoning their previous party—either motivated by affect of powerlessness—as
well as drawing in previous non–voters, resulting in a substantial shift. The new dynamic may
only last for an election, where the exit option is a sort of pressure value during the election,
with voters returning to their original party in the next election; alternately, the party may
sustain, more permanently altering elections in a country. The Pim Fortuyn Lijst could be a
case of such a scenario. A few months after inception, the party had rapid success, placing third
in the national elections. However, after the assassination of Pim Fortuyn, the party quickly
collapsed nearly as quickly as it rose, losing a large portion of its vote share in the following
elections. While the number of political parties has been fairly stable, the appearance of a new
alternative had significant implications for the 2002 elections. While there were other outsider
parties, their support levels were fairly stable; before LPF, there was volatility, but not to the
same extent as the 2002 parliamentary elections that created a shock, quickly increasing levels
of vote switching and decreasing mainstream party support.
6.5
A Protest Voter?
The current conceptualization of a protest voter is an individual who non–habitually votes
with the intention of registering dissatisfaction without regard for ideology and other political
attitudes (Van der Brug & Fennema 2003, Van der Brug & Fennema 2007). The theorization
lacks clarity and establishes a strange operationalization. The theory seems to treat a protest
vote as an entirely separate phenomenon from demonstration attendance or other political
actions. Social movements and party choice—or the decision to case a protest vote—are treated
as unique actions with separate motivations. While there is not a clear operationalization of a
protest voter, a social movement approach may have potential to explain the possible presence
of protest voting. Based in the attitudes, motivations, and goals of attending demonstrations
or undertaking some other form of protest, the protest voter may be defined as the institutional
expression of the factors behind non–institutional protest. As unconventional behaviors may
form a common repertoire, protest could be incorporated, in which there is an institutional and
extra–system manifestation, with both influenced by the same factors. Given the role of internal
72
and external efficacy in non–institutional protest (Gamson 1968, Craig 1980), both attitudes
would also be expected to maintain relevance in the institutional form of protest. Using efficacy
and satisfaction can more fully explain, and provide a definition of protest voting, as the two
variables complement each other to explain protest attendance.
While ideology may not be the sole motivating force of unconventional behaviors, it
cannot be discounted in untraditional behaviors, and goes against the body of work on spatial
voting. While the analysis supports a role for attitudinal variables in non–conventional voting,
the results also support the influence of ideological factors. Instrumental motivations alone
may not give a full understanding, but a purely–expressive framework is unable to broadly explain behavior, as well. Under the current operationalization, when deciding to cast a punitive
vote, it is unlikely the ideology would have no influence. Either acting spitefully and selecting
an oppositional party to one’s previous party’s position, or choosing an ideologically similar
one in a signaling–like fashion, ideological positioning is expected to have some form of influence. Likewise, the motivation to protest could also come for the dissatisfaction from policy
disapproval. The 1998 Swedish elections, in which there were low levels of trust in the two
mainstream parties, saw a large amount of voting switching to peripheral parties. Voters used
available exit options, motivated by discontent, as well as ideology (Moller 1999), impressing
on the importance that ideology can have in an affective response. And when examining social
movements, there are underlying grievances and ideological considerations that can motivate
an individual to action.
7
Conclusion
There is a connection between the motivations for non–traditional behaviors outside the political system and non–traditional behaviors within the system; the drivers of social movements
and extra–system participation have influence in institutional political action. Those that are
disaffected with government and feel politically powerless are more likely to participate through
means other than the ballot box, but what of the alienated and dissatisfied that still vote? There
is a portion of the electorate that is inefficacious, but, despite their inefficacy, they still engage
with the system. However, they do not behave traditionally: an inefficacious voter is more likely
to change their party loyalty and support outsider movements. The alienated and dissatisfied
73
do not necessarily behave in the standard ideologically–motivated actions. Both social movements and electoral behavior are similar phenomena, comprising a repertoire of unconventional
behaviors, in which the individual is motivated by disaffection and alienation, and their goal
is to restore their political influence and positive affect. An individual satisfied and empowered will make conventional political decisions, correctly voting based on ideological differences
between themselves and parties, and emphasizing instramental considerations. However, the
analysis suggests that ideology may be a second–order consideration. Before engaging in traditional electoral decision–making, an individual must feel a sense of empowerment and have a
sense of satisfaction with the government. If these two conditions are not met, the individual
will attempt to resolve their powerlessness or dissatisfaction; the power dynamics and affect
towards government become the motivating factors behind their choices, transitioning ideology
to a second–order consideration. The individual becomes more expressively motivated, in which
idealogical concordance between themselves and a party are de–emphasized, and restoring their
government influence and affect become more prominent in decision making. While maximizing
influence is not mutually exclusive from spatial considerations, the alienated will be motivated
by the former. When faced with the choice between equally–attractive parties, ideology alone
cannot discern which party the individual will choose. Likewise, the use of ideology alone cannot
inform the point at which one will abandon their current party in favor of another, or when a
voter will choose to support an outsider political party.
7.1
Future Work
While non–traditional, extra–system and institutionally–based participation have similar motivating characteristics, it remains to be seen whether the two forms of behavior have greater
interconnections. The two may be influenced by similar factors, but that does not necessarily
imply that two have further similarities or that the two forms behaviors play off one another.
Does to inefficacious voter cast a ballot to send signals or register discontent before using other
means? When faced with continued unsuccessful action at the ballot box, does the individual
then transition towards non–institutional behavior? Blank and null voting can be a predictor for
future political unrest; can electoral volatility and outsider party support be such a predictor, as
well? Future work should examine the extent to which the institutional and non–institutional
repertoires of unconventional action intermingle. The predictability of one form of action for
74
another, as well as the prominence of engaging in both electoral and extra–system mean provide
future paths to understanding the repertoire.
The analysis suggests that democratic deficits are relevant factors in political decisions,
but what are the roots of political powerlessness? When faced with parties continually pursuing
policies unfavorable to the individual, it is possible for feelings of powerlessness to emerge, but
effects individuals differently. Although not explored in this paper, it would be worthwhile for
future work to examine the implications of Craig’s (1980) notion of entitlement. A possible
operationalization of democratic deficits or political powerlessness, could be created from the
entitlement understanding. The roots of inefficacy, perceptions of one’s inability to influence
the system, could arise from a disconnect between the preferred and actual amount of influence.
The internally efficacious fell that they are capable enough to hold a large amount of influence
within the system, while the externally inefficacious feel that the actual means to influence are
limited. This gap could be to origin of democratic deficits. Likewise, an operationalization of
democratic deficits could employ internal efficacy: the separation of ones perceived competency
to influence government and the government’s responsiveness. Those that see themselves as
competent enough to create their desired change, but are unable to influence an unresponsive
government, would have a greater sense of a deficit.
A next step would be the analysis of the generalizability of an efficacy–ideology model,
testing the social movements approach in different systems, as well as across time with panel
data. Cross–country testing would enable a better understanding of efficacy’s influence on voting
across different political histories and institutions. Electoral districts with variable magnitudes
and electoral thresholds can modify individuals behavior, as well as the effects of efficacy on
vote choice. The implications of political powerlessness may have variability across democracies,
such as the newer democratic systems in Eastern Europe or First Past The Post Systems like
the United States’. The model could also be applied to further the understanding of signaling, political breakthroughs, as well as the importance of symbolic citizen control outlined in
Coleman and Davis’ work (1976). Time series analysis will allow for a more in–depth understanding of the point at which an individual will change their party support or transition to
an outsider political party, as well as when an individual will transition between institutional
and extra–system means of influence. While the analysis shows a relation between dissatisfaction and inefficacy with these behaviors, it is less beneficial for informing at which level of
75
alienation or dissatisfaction one will change. Additionally, data over time will give insight into
whether restorative actions are effective in re–creating political empowerment or satisfaction
for the individual. Likewise, time series will be able to inform whether mainstream parties are
able to re–capture voters that have moved to outsider parties. The data could give insight to
the effectiveness of restorative actions, as well as the effectiveness of political parties’ strategies
for re–gaining voters. Over–time data would allow for an analysis of the reciprocity between
restorative action and efficacy and satisfaction levels—whether an individuals affect becomes
more positive and they become more empowered, or restorative action is ineffective, or if there
is a spiral of distrust.
The use of an efficacy and left–right ideology model could be beneficial, also though
it would lose the more in–depth aspect of individual policy areas, but reduces some drawbacks.
Salient policy topics lose and gain salience over time, which presents issues for analysis. While
individuals’ policy positions may change over time, democratic ideals and voters’ optimal level of
influence are more fixed—perceptions of how government is upholding those ideals and deficits
are what fluctuate. Predicting which policy issue will be the most important for a given election
can be difficult. And more contentious policy areas such as immigration or euthanasia are more
subject to social desirability pressures or respondents refusing to answer, which decreases the
total amount of available data. An attitudinal–ideological framework of voting behavior has
potential for a more full understanding of voters’ decision making; however, the analysis raises
more questions and possible implications. The model provides a framework connecting social
movements with electoral behavior, analyzing contexts in which voters are motivated by other
factors than ideology. Future analyses can tease out the relative importance and effects of
efficacy and satisfactions, as well as test the interactions between the two. More than how a
voter responds to alienation, future work should also focus on how political parties react in such
environments, and whether their strategies are successful. Given the introduction of a new exit
option, there is potential for increased volatility, but will the increase fluidity in elections act
as a pressure valve, quickly returning to normal levels, or serve as a more extended flux?
The introduction of efficacy and satisfaction gives additional insight into electoral
decision–making. Factors influencing social movements play a part in vote choice; and there
are situations in which individuals’ left–right positioning is not the main consideration. As
opposed to dichotomize institutional and extra–system action, integrating the two behaviors
76
can provide a greater understanding of individuals’ political actions. It is worth reconsidering
the preeminence of ideology and the assumption that voters’ behavior is spatially–motivated.
The inefficacious electorate’s choices de–emphasize spatial considerations, implying that they
do not engage in correct voting; however, given their goals, they cannot be said to incorrectly
vote. Where correctly voting is defined in terms of supporting the party with which on agrees
most ideologically, analysis of voting that best fulfills a voter’s intention may provide additional
understanding; the factors that affect an individual’s goal of casting a ballot—such as when one
desires to minimize ideological differences, promote their influence within an organization, or
have the greatest effects on policy outputs—can provide a basis for understanding citizens voting
strategies, as well as the roots of increasing electoral instability and the declining mainstream.
77
8
Appendix
8.1
Appendix A: Variables
Internal Efficacy
Although internal efficacy is employed in the protest literature, it is not used in this paper’s
framework. This project’s focus is on the power relationship between citizen’s, parties, and government, as well as dissatisfaction. The attitudinal object of interest is the state—the extent to
which the individual feels that the government is willing and able to implement their desires—
and less so one self competency captured by internal efficacy. One’s perception of their ability
to influence and understanding government is not foundational to restorative action: both the
internally efficacious and inefficacious can hold attitudes towards democratic deficits, and desire
to correct the deficits. The internally inefficacious are faced with a meaninglessness alienation:
“the individual is unclear as to what he ought to believe—when the individual’s minimal standards for clarity in decision–making are not more” (Seligson 1959, 786). This alienation caused
by perceptions of over–complexity and a lack of understanding has the potential to motivate
actions to regain a sense of clarity, which could have implications for vote choice, but is out of
the scope of this particular analysis.
In future work, it is worthwhile to examine the relationship between internal and external efficacy, how they influence voting behavior, and a possible operationalization of democratic
deficits. Craig’s interpretation of Gamson’s hypothesis includes an aspect of the internally efficacious feeling more capable, and therefore have an entitlement–like desire for more influence
(1980). This desire for greater influence, due to individuals’ perceptions of self–competency,
may manifest in a sense of democratic deficits. The magnitude of the discrepancy between an
individual’s perceived ability to influence and the government’s responsiveness may inform the
extent to which an individual senses that there are democratic deficits.
DPES Surveys
Given that some variables were dropped in some studies, not every election from 1971–2006 can
be analyzed. The total number of complete cases is N = 6, 845. The first two, 1982, and 2003
elections were not included in the analysis, as one or more of the variables were not included in
the survey. Complete data are more prevalent in the later, post 1980 election studies.
Year
Cases
1971
0
1972
0
1977
486
1981
817
1982
0
1986
720
1989
999
1994
782
1998
996
2002
956
2003
0
2006
1089
DPES Survey Question Language
Vote Intention Which party do you intend to vote for?
Asked if respondent answered “yes” to: As you may know, elections for the Second Chamber will be held in [Month/Date] of this year. Do you intend to vote or not, or dont you
know yet?
78
Previous Party Choice For which party did you vote then?
Asked if respondent answered “yes” to: The previous elections for the Second Chamber
were held in [year of previous election]. Did you vote in these elections, or not?
External Efficacy This survey topic contains five yes/no questions:
1. MPs do not care about opinions of people like me
2. Parties only interested in my vote, not my opinion
3. People like me have no influence on politics
4. So many people vote. my vote does not matter.
Not included in the 2006 study.
5. MPs quickly lose contact with citizens.
Only in the 1998, 2002, and 2006 studies.
Government Satisfaction With the help of this card, could you indicate how satisfied you
are in general with what the government has done during the past four years?
One is “very satisfied” and five is “very unsatisfied.”
Class One sometimes speaks of the existence of various social classes and groups. If you were
to assign yourself to a particular social class, which one would that be?
One is upper class, five is working class.
Income The respondent has been asked to indicate, with the help of categories on a showcard,
the level of household income. The interviewer emphasized that the information provided
would remain strictly confidential and that the question referred to the net total income
of the household (i.e. the sum of net incomes of all members of the household, including
social security, unemployment benefits, etc., after deduction of taxes).
Education The next question is about your own education. Could you indicate by means of
this showcard the highest education for which you received a diploma?
One is elementary, five is higher level vocational, university.
Religion How often do you attend religious services?
One is at least weekly, five is almost never or never.
Political Interest Are you very interested in political subjects, fairly interested or not interested?
Policy Position Question Wording
Abortion Voluntary ending of pregnancy. Some people think that the government
should forbid abortion in all circumstances, other people think that each woman
should have the right to decide for herself whether or not she wants an abortion. Of
course there are also people who have an intermediate opinion. Suppose we place
the persons who would like to forbid in all circumstances at the left of this line, at
number 1, and the persons who think each woman has the right to decide for herself
at the other end, at number 7.
Asylum Now I would like to talk with you about another problem. Allowing ASYLUM
SEEKERS to enter the Netherlands has frequently been in the news during the last
few years. Some people think that the Netherlands should allow more asylum seekers
than the government currently does. Other people think that the Netherlands should
79
send asylum seekers who are already staying here back to their country of origin.
Of course, there are also people whose opinion lies somewhere in between. At the
beginning of this line are the people (and parties) who think that the Netherlands
should allow more asylum seekers to enter (at number 1); at the end of the line are
the people (and parties) who think that the Netherlands should send back as many
asylum seekers as possible (at number 7).
Minorities There is disagreement in the Netherlands about foreigners and ETHNIC
MINORITIES. Some people and parties think that these people should be able to
live in the Netherlands while preserving all customs of their own culture. Others
think that these people, if they stay in the Netherlands, should completely adjust
themselves to Dutch culture. Of course, there are also people whose opinion is
somewhere in between. At the beginning of this line are the people (and parties) who
think that foreigners and ethnic minorities should be able to live in the Netherlands
while preserving all customs of their own culture (at number 1); at the end of the
line are the people (and parties) who think that these people should fully adjust
themselves to Dutch culture (at number 7).
Income Inequality Some people and parties think that the differences in incomes in
our country should be increased (at number 1). Others think that these differences
should be decreased (at number 7). Of course, there are also people whose opinion
is somewhere in between.
EU Integration EUROPEAN UNIFICATION is well under way. The countries of the
European Union have decided to work more and more closely together. However,
not everybody holds the same view about it. Some people and parties think that
European unification should go further. Others think that European unification
has already gone too far. Suppose the people and parties who think that European
unification should go further are at the beginning of this line (at number 1) and the
people and parties who think European unification has already gone too far are at
the end of the line (at number 7).
Euthanasia Now some questions about political affairs that are frequently in the news.
When a doctor ends the life of a person at the latters request, this is called euthanasia. Some people think that euthanasia should be forbidden by law. Others feel that
a doctor should always be allowed to end a life, if the patient makes that request.
Of course, there are people whose opinions lie somewhere in between. Suppose that
the people (and parties) who think that euthanasia should be forbidden are at the
beginning of this line (at number 1) and the people (and parties) who feel that a
doctor should always be allowed to end a life upon a patient’s request are at the
end of the line (at number 7)
Nuclear Power Now I would like to ask you some questions about political issues that
are regularly in the news. As you may know, some people fear that within the
foreseeable future a shortage of energy will occur in the world. One means of
supplying this shortage is to build NUCLEAR POWER PLANTS. Some people
therefore believe that the Netherlands should quickly increase the number of nuclear
power plants. Others, on the other hand, consider the dangers too great and think
that no nuclear power plants should be built at all. At the beginning of this line
are people and parties who think that additional nuclear plants should be built in
the Netherlands (at number 1); at the end of this line are people and parties who
think that no nuclear plants should be built at all (at number 7).
80
Table 3: Variables by study Year
Abortion
1971
1972
1977
1981
1982
1986
1989
1994
1998
2002
2003
2006
Asylum
Minorities
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Income Ineq.
x
x
x
x
x
x
x
81
EU
x
x
x
x
Euthanasia
x
x
x
x
x
x
x
Nuclear
Left–Right
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Post–wave
x
Total
0
0
3
3
3
5
5
6
7
7
5
7
8.2
Appendix B: Vote Switching
Table 4: Changing Party Loyalties: Intension and Actual
PvdA CDA VVD D66 GL SGP CU LPF SP Other Not Vote
PvdA
2529
103
21
85
19
1
2
3
36
57
112
CDA
71
2749
86
32
1
5
21
3
10
33
80
VVD
22
109
1423
23
3
0
1
8
5
23
51
D66
65
26
46
588
10
1
0
1
4
19
30
GL
54
8
11
27
326
0
0
2
18
10
6
SGP
1
10
0
1
0
152
11
0
1
1
4
CU
0
29
3
0
0
3
249
0
0
2
2
LPF
1
10
6
0
0
0
0
100
2
1
3
SP
32
5
4
6
10
0
0
1
270
8
2
Other
62
39
23
27
10
2
6
4
10
342
25
Not Vote
230
206
101
86
28
5
9
6
58
96
746
Undecided
320
429
244
164
48
9
43
1
148
179
80
Total
3387
3723 1968 1039 455 178 342 129 562
771
1141
Separate from table 2, this table is a voter’s intended party they will support, asked during the
pre–election wave, contrasted with the self–reported party they supported, asked during the
post–election wave.
Total
2968
3091
1668
790
462
181
288
123
338
550
1571
1665
Policy Distance
Table 5: Vote Switching Regression Table with Policy Distance
Intercept
Efficacy
Satisfaction
Policy
Income
Class
Education
Age
Mean
0.328
0.629
0.660
0.624
0.876
1.168
1.495
0.595
Lower CI
0.296
0.551
0.589
0.504
0.772
1.004
1.288
0.520
Upper CI
0.364
0.715
0.743
0.774
1.002
1.362
1.727
0.680
Effective Size
9900.000
8301.000
9130.000
9381.000
9776.000
9101.000
9574.000
9300.000
The model is the same as the vote switching model, except that the absolute value
of the policy preferences is used. Instead of measuring general right–left placement in the
policy space, the variable measure the relative extremeness of an individual, as compared
to the mean. The results are very similar to the regression with general policy preferences,
with the only difference being that the policy variables has a negative relationship. As one
becomes relativly more liberal or conservative in their policy preferences, they become less
likely to change their party loyalty. Given that an individual is relativly more extreme
than the mainstream parties, there are fewer alternatives or exit options available to
them; the individual will support a more niche party that captures a unique policy space,
with which they agree.
82
Under–Prediction of Vote Switching from the Median
Let X ∼ Binomial(n, θ), where n is the number of simulations and θ is the PPD of the
Pn/2−1 R
proportion of vote switchers. P (Y = 1) = 1 − P (X < n/2) = 1 − k=0 nk θk (1 −
θ)n−k dθ. E[X] = E[X | θ = θ̂] = nθ̂. Therefore, θ ≥ .5 for the median to be one. Since
E[θ] = .27 and has a relativly narrow variance, it is highly unlikely that an observation’s
simulation median drawn from the PPD will be equal to one, despite the expected 27%
of vote switchers.
Hesitation With Regards To Party Choice
Table 6: Hesitated Party Choice
Intercept
Efficacy
Satisfaction
Policy
Income
Class
Education
Age
Mean
0.331
0.783
0.781
0.917
0.967
1.181
1.722
0.417
Lower CI
0.309
0.687
0.689
0.799
0.847
1.010
1.486
0.360
Upper CI
0.354
0.896
0.884
1.064
1.106
1.387
2.011
0.481
Effective Size
9453.000
9439.000
9431.000
9355.000
9107.000
8952.000
9345.000
8816.000
Table 6 are the results of an individual’s hesitation to support their chosen party,
regressed on the vote switching model. A one indicates a yes—zero a no—to the question
“And did you seriously consider voting for [a different party than the one intended] at the
parliamentary elections?” The results support the findings of the vote switching model:
the inefficacious and dissatisfied are more likely to consider alternative parties. While
not all act, the alienated and disaffected contemplate changing their party loyalty more
so than the empowered and satisfied. Unlike the vote choice model, policy position is
less likely to be a factor in party choice hesitation; individuals across the policy space
have relativly equal amounts of hesitation. Policy preferences have less impact on party
hesitation, while individual’s attitudes are more influential.
83
8.3
Appendix C: Mainstream Voting
Relative Ideological Extremeness
Table 7: Mainstream Voting Regression Table with Left–Right Relative Extremeness
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
Mean
3.656
1.294
2.263
0.625
1.399
0.919
0.405
1.863
Lower CI
3.267
1.148
2.030
0.510
1.238
0.800
0.352
1.637
Upper CI
4.096
1.467
2.528
0.771
1.590
1.063
0.467
2.117
Effective Size
9061.000
8594.000
9454.000
9180.000
9248.000
9041.000
9289.000
9083.000
Left–right distance is measured as the absolute value of an individual’s left–right
self–placement. As the mean is zero, the variables reflects how far outside the mean
an individual places themselves ideologically. The other regressors’ effects are virtually
unchanged, while the left–right extremeness variable’s relation is negative. Those that
are more conservative are more likely to support the mainstream, but those that are
further from the mean ideology are less likely to support the mainstream. As two of
the three mainstream parties are conservative, it follows that the mainstream attracts
more conservative voters; likewise, as the PvdA, CDA, and VVD are more moderate,
individuals’ whose ideology is relativly more far left or right tend to support outsider
parties.
Mainstream Voting with Left and Right Categorization
Table 8: Mainstream Voting Among Liberal Parties
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
PCC
Mean
1.790
1.331
1.227
0.698
1.201
0.473
0.307
1.615
.6605
Lower CI
1.610
1.113
1.049
0.567
1.005
0.389
0.252
1.345
Null
Upper CI
2.010
1.580
1.427
0.869
1.421
0.577
0.376
1.950
.6152
Effective Size
9054.000
9102.000
9227.000
9052.000
8890.000
8884.000
9330.000
8713.000
Table 8 are the results from logistic regression, in which a one represent an
intended vote for the PvdA, and a zero indicates an intended vote for GroenLinks, D66,
or the Socialist Party. Table 9 are the results from logistic regression, in which a one
represent an intended vote for the VVD or CDA, and a zero indicates an intended vote
for ChristenUnie, LPF, or the SGP.
The decision to support the mainstream left party over the outsider left is influenced is positively associated with efficacy and satisfaction. As an individual becomes
84
Table 9: Mainstream Voting Among Conservative Parties
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
PCC
Mean
9.590
1.132
3.361
0.340
1.525
1.607
0.506
1.510
.8784
Lower CI
8.023
0.890
2.684
0.249
1.216
1.214
0.387
1.190
Null
Upper CI
11.454
1.430
4.220
0.460
1.977
2.140
0.667
1.919
.8790
Effective Size
8995.000
8832.000
9128.000
9045.000
8771.000
9148.000
9169.000
8857.000
more empowered and or more satisfied, the probability that they will support the more
established liberal party—the PvdA—increases. Given the data, there is a 99.93% chance
that efficacy in non–negative and positively related to supporting the PvdA over the leftist
outsiders; for satisfaction, the probability is 99.57%. Deciding to support one of the two
mainstream conservative parties has an almost surely non–zero and positive relation with
satisfaction; however, the probability that efficacy has a positive effect is 84.33%. While
the HPD region of the efficacy coefficient contains zero, there is still a fairly high probability that a more efficacious voter will support the mainstream right over the outsider
conservative parties.
Mainstream Voting by Election
When examining individual elections, the results are fairly close to the effects of attitudes
and ideology on mainstream support across all elections. The probability that efficacy is
positively related to mainstream vote intention is greater than 95% for the 1977, 1981, and
2002; 89% for 1994; 86% for 2006; dropping to 76% for the 1998 elections; and it is unlikely
efficacy had an effect for the 1986 and 1989 elections. Satisfaction is more consistent,
with a greater than 95% probability of a positive relationship for all elections, except
for 1986, in which the probability is 85%. Left–right self–placement is fairly consistently
positively related to mainstream voting, suggesting that conservative respondents are
more likely to support the mainstream, with probability greater than 95% for all elections
besides 1986 and 1989, in which there is likely no effect of left–right ideology. Differing
dynamics of individual elections have the potential to alter the factors behind mainstream
support, deviating from the broader trend. In particular, respondents support for the
mainstream during the 1986 survey are mostly unaffected by attitudinal factors, as well
as traditional ideological components. Efficacy levels and self–placement of the liberal–
conservative scale were likely non–factors in mainstream party support, but satisfaction
with government retained an 85% probability of affecting the decision–making process.
As would be expected, attitudinal and ideological factors were significant in the 2002
elections that swept LPF to power. Comparing the efficacy of the 2002 respondents to
others, theirs was th lowest of all; however, they were slightly more satisfied than average
and tended to be more liberal.
85
Table 10: Mainstream Voting by Election, 1
1977
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
N
Mean
4.516
1.783
1.873
1.964
1.072
1.588
0.287
1.781
486
Lower CI
3.235
0.937
1.042
1.179
0.474
0.772
0.138
0.977
Upper CI
6.310
3.326
3.328
3.252
2.285
3.280
0.593
3.176
Effective Size
6701.000
6846.000
6657.000
6428.000
6911.000
6978.000
6384.000
6833.000
Upper CI
4.323
2.051
2.671
2.359
1.746
1.123
1.114
3.876
Effective Size
6625.000
6578.000
6775.000
6791.000
6739.000
6893.000
7087.000
6718.000
Upper CI
9.021
1.784
2.085
1.578
1.542
2.074
1.076
3.016
Effective Size
6480.000
6738.000
6702.000
6741.000
6938.000
7004.000
6846.000
6674.000
Upper CI
4.891
1.488
3.067
1.557
1.873
1.163
0.966
3.824
Effective Size
6809.000
6787.000
6973.000
6769.000
6751.000
6867.000
6827.000
7134.000
1981
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
N
Mean
3.522
1.401
1.880
1.626
1.157
0.746
0.723
2.566
817
Lower CI
2.887
0.977
1.323
1.102
0.798
0.491
0.458
1.737
1986
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
N
Mean
7.030
1.052
1.299
0.928
0.938
1.187
0.616
1.771
720
Lower CI
5.545
0.639
0.785
0.555
0.564
0.681
0.361
1.026
1989
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
N
Mean
4.090
1.012
2.126
1.074
1.324
0.777
0.595
2.601
999
Lower CI
3.405
0.696
1.462
0.730
0.918
0.516
0.365
1.767
86
Table 11: Mainstream Voting by Election, 2
1994
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
N
Mean
2.078
1.291
2.881
2.421
1.684
0.907
0.865
3.375
782
Lower CI
1.717
0.862
1.980
1.746
1.196
0.599
0.525
2.245
Upper CI
2.500
1.908
4.205
3.400
2.321
1.364
1.417
4.930
Effective Size
6859.000
6973.000
6802.000
6621.000
6753.000
6849.000
6964.000
6953.000
Upper CI
3.090
1.642
4.839
3.037
2.232
1.012
0.591
3.111
Effective Size
6701.000
6671.000
6621.000
6567.000
6541.000
6683.000
7185.000
6547.000
Upper CI
2.213
1.897
5.586
3.859
3.228
0.968
0.971
2.936
Effective Size
6741.000
6649.000
6711.000
6722.000
6739.000
6596.000
6548.000
6997.000
Upper CI
3.106
1.524
3.040
3.116
2.012
1.593
0.884
3.210
Effective Size
6788.000
6905.000
6752.000
6791.000
6843.000
7109.000
6931.000
6676.000
1998
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
N
Mean
2.534
1.143
3.358
2.192
1.593
0.666
0.388
2.149
996
Lower CI
2.097
0.793
2.307
1.551
1.136
0.423
0.259
1.481
2002
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
N
Mean
1.767
1.360
4.019
2.777
2.231
0.656
0.674
2.064
956
Lower CI
1.434
0.950
2.873
1.992
1.545
0.438
0.464
1.471
2006
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
N
Mean
2.523
1.168
2.254
2.281
1.470
1.120
0.586
2.316
1089
Lower CI
2.036
0.899
1.663
1.661
1.077
0.778
0.376
1.624
87
Percent Mainstream Vote by Election
Election
1977 1981 1986 1989 1994 1998 2002 2006 Average
Percent Mainstream Vote
86
75
87
80
66
71
59
69
73
Average Efficacy Score
.12
.11
.12
.14
.15
.04
-.01
.06
.09
Average Satisfaction Score
.17
-.10
.02
.04
-.10
.20
.04
.01
.03
Average Left–Right Score
.04
.06
.00
.02
.02
-.05
-.09
.00
.00
The prior for each regression is a Cauchy distribution, with th MLE from frequentist regression serving
as the location parameter, and the scale set to 2.5. There was a 50,000 iteration burnin, with an
additional 200,000 iterations thereafter, with every 10th iteration stored. The acceptance rates were
approximately 15% for each model.
Although mainstream support in 1986 was likely unaffected by efficacy, it may
have been a sign of things to come. There was a shift in the electoral dynamics that did
not manifest in large outsider party gains, but there were large number of vote switching,
with the CDA and PvdA making unexpected gains, and the VVD taking large losses (Van
Der Eijk et al. 1986). The 1986 elections proceeded the decline in mainstream support,
in which the respondents in future studies indicated support for PvdA, CDA, and VVD
in lower numbers; after 1986, all following DPES studies had a lower vote share for the
mainstream. The 1986 election may be a case in which large numbers of voters switch
parties as a form of restorative action, and when it was unsuccessful, turned to outsider
parties. There may have been a progression in the unconventional repertoire, in which
larger numbers of the Dutch electorate engage in untraditional voting behaviors follow
the 1980s.
Posterior Predictive Diagnostic
Table 12 is the instances of misclassification between the PPD and the observed result,
regressed on the model’s variables. The regression gives information of the linear combination of predictors that tend to produce either a false positive or a false negative in the
simulations of the PPD. That is, for a false positive, what regressors are most likely to
predict an individual will vote for a mainstream party when they did not; for a false negative, which variables are related to falsely predicting an individual will not support PvdA,
CDA, or VVD when they do. Logistic regression was used, and the coefficients are in
log odds. An increase in efficacy, satisfaction, and left–right placement decrease the odds
of a false positive; each variable increases the odds of false negatives, with satisfaction
the largest effect, ideology the next, and efficacy the least. In 99% of simulations from
the PPD, these variables will be significantly related to misclassification. The regression
suggests that the inefficacious, unsatisfied, and liberal individuals are the largest source
of false positives. Figure 10 depicts the differences between efficacy and satisfaction of
liberal and conservative voters. Ideology was dichotomized into more left–leaning than
average—a self–placement less than 0—and more conservative than average. Likewise,
satisfaction and efficacy were divided into more and less satisfied or efficacious than the
average. The differences between the proportion of above–average levels of dissatisfaction
and inefficacy between liberal and conservative voters is examined with Fisher’s exact test
for contingency tables in a Bayesian context. The posterior distributions of the conser88
Figure 9: Posterior Distribution of Regression Coefficients
7
5
4
3
2
0
1
1
0
0.0
0.5
1.0
1.5
−0.5
0.5
1.0
1.5
2.0
Log Odds
Left−Right
Efficacy, Satisfaction, and Left−Right
8
Log Odds
Efficacy
Satisfaction
Left−Right
0
0
1
2
2
4
Density
6
All Years
1977
1981
1986
1989
1994
1998
2002
2006
3
4
5
6
7
−1.0 −0.5
Density
All Years
1977
1981
1986
1989
1994
1998
2002
2006
6
All Years
1977
1981
1986
1989
1994
1998
2002
2006
Density
3
Satisfaction
2
Density
4
5
6
Efficacy
−0.5
0.0
0.5
1.0
1.5
0.0
Log Odds
0.2
0.4
0.6
0.8
1.0
Log Odds
vative proportion, subtracting the liberal proportion, appear displayed in 10. As with
the χ2 test of independence, a value of zero indicates that the two are the same. The
distribution for the difference density of the inefficacy proportions is centered around
zero; therefore, conservative and liberal voters most probably have the same rates of inefficacious voters. For dissatisfaction, the distribution is negative, and does not include
zero. Liberal voters almost surely have a higher rate of above–average dissatisfaction
than conservative voters.
In the Bayesian framework, the test is constructed as follows:
y1 ∼ Binomail(n1 , θ1 ) and y2 ∼ Binomial(n2 , θ2 )
A uniform prior is used for both parameters, which is a Beta(1, 1) distribution.
From the conjugacy of the Beta Binomial distribution,
89
Table 12: Posterior Predictive Distribution Misclassification
False Positive
False Negative
(1)
(2)
−0.203∗∗∗
(0.061)
−0.235∗∗∗
(0.056)
−0.231∗∗∗
(0.057)
0.030
(0.060)
0.023
(0.070)
0.315∗∗∗
(0.069)
−0.372∗∗∗
(0.061)
−1.212∗∗∗
(0.029)
0.431∗∗∗
(0.062)
1.184∗∗∗
(0.058)
0.747∗∗∗
(0.059)
0.544∗∗∗
(0.061)
−0.300∗∗∗
(0.071)
−1.104∗∗∗
(0.072)
0.925∗∗∗
(0.063)
1.004∗∗∗
(0.031)
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
Constant
∗p
< .1;
∗∗ p
< .05;
∗∗∗ p
< .01
Figure 10: Efficacy and Satisfaction of Left and Right–Leaning Voters
Efficacy and Satisfaction of Left and Right Voters
15
10
5
0
Density
20
25
30
Inefficacy
Dissatisfaction
−0.2
−0.1
0.0
0.1
Difference in Proportions
90
0.2
p(θ1 | y1 , n1 ) = Beta(θ1 | y1 +1, n1 −y1 +1) and p(θ2 | y2 , n2 ) = Beta(θ2 | y2 +1, n2 −y2 +1)
The posterior distribution is sampled from ϑ = θ1 − θ2 , giving the convolution:
Z
p(ϑ | y, n) = Beta(θ | y1 + 1, n1 − y1 + 1)Beta(θ − ϑ | y2 + 1, n2 − y2 + 1)dθ
Where the Beta distribution has the following form, as y and n are both integers:
Z
Γ(α + β) α−1
f (x; n, y) =
x (1 − x)β−1 , α = y + 1 β = y − n + 1
Γ(α)Γ(β)
The probability that the difference of proportions is not zero is given by p(ϑ > 0). Figure
10 is the estimated density of p(ϑ), using 100,000 samples.
Attitudinal–Ideology Compared to Ideology–Only Model
The model of mainstream voting with efficacy, satisfaction, and ideology outperforms
the model with only left–right positioning. While the attitudinal model has a PCC of
.7315965, the PCC for an ideology–only model is .7278557. The pre ideological model
also increases the rate of false–positives at 26.37%, which over–predicts mainstream voting
more so than the attitude–ideology model, with a 25.4% false positive rate. Differences
in the classification ability can be seen in the ROC Cure figure. When comparing two
frequentist models, the AIC for the ideology–only model is 7510.844, while the attitude–
ideology model’s is lower, with an AIC of 7317.934. The pattern is the same for the
Bayesian Information Criteria, with the ideology’s BIC 7551.832, as compared to the
combined model’s at 7372.584.
1.0
ROC Curves
0.6
0.4
0.2
0.0
True Positive Rate
0.8
ID and Attitudes
Attitude Only
ID Only
0.0
0.2
0.4
0.6
False Positive Rate
91
0.8
1.0
8.4
Appendix D: Frequentist Regression Results
Table 13: Frequentist Regression Tables
(1)
(2)
−0.456
(0.066)
−0.398∗∗∗
(0.059)
0.150∗∗
(0.065)
–
Satisfaction
Policy
Left–Right
Income
Class
Education
Age
Constant
N
Log Likelihood
AIC
p < .1;
Vote Mainstream
∗∗∗
Efficacy
∗
Vote Switch
∗∗
−0.128∗
(0.066)
0.143∗
(0.077)
0.446∗∗∗
(0.074)
−0.530∗∗∗
(0.069)
−1.292∗∗∗
(0.032)
6,845
−3,476.510
6,969.021
p < .05;
∗∗∗
p < .01
92
0.290∗∗∗
(0.064)
0.717∗∗∗
(0.058)
–
0.559∗∗∗
(0.060)
0.316∗∗∗
(0.064)
−0.171∗∗
(0.074)
−0.787∗∗∗
(0.073)
0.591∗∗∗
(0.066)
1.107∗∗∗
(0.032)
6,845
−3,650.967
7,317.934
Vote Choice Frequentist Regression
Satisfaction
Left–Right
Policy
Income
Class
Education
Age
Religion
Constant
AIC
Satisfaction
Left–Right
Policy
Income
Class
Education
Age
Religion
Constant
AIC
∗
p < .1;
∗∗
VVD
Undecided
(1)
(2)
(3)
0.105
(0.232)
−0.068
(0.089)
0.864∗∗∗
(0.043)
−0.256∗∗∗
(0.065)
0.167∗∗∗
(0.026)
−0.783∗∗∗
(0.088)
0.261∗∗∗
(0.060)
−0.005
(0.005)
0.186∗∗∗
(0.054)
1.857∗∗∗
(0.283)
6,253.714
0.403∗
(0.212)
0.133∗
(0.079)
0.444∗∗∗
(0.036)
−0.087
(0.059)
0.075∗∗∗
(0.024)
−0.219∗∗∗
(0.075)
0.274∗∗∗
(0.055)
−0.010∗∗
(0.004)
−0.171∗∗∗
(0.047)
0.797∗∗∗
(0.253)
6,253.714
∗∗
Efficacy
Efficacy
CDA
0.400
(0.191)
−0.310∗∗∗
(0.074)
0.654∗∗∗
(0.033)
−0.324∗∗∗
(0.053)
0.071∗∗∗
(0.022)
−0.278∗∗∗
(0.067)
0.209∗∗∗
(0.049)
0.015∗∗∗
(0.004)
−0.444∗∗∗
(0.041)
1.565∗∗∗
(0.228)
6,253.714
GroenLinks
SGP
ChristenUnie
LPF
SP
Other
Undecided
(1)
(2)
(3)
(4)
(5)
(6)
(7)
0.143
(0.442)
−0.080
(0.163)
−0.564∗∗∗
(0.084)
−0.369∗∗∗
(0.137)
−0.103∗∗
(0.045)
−0.181
(0.163)
0.262∗∗
(0.124)
0.018∗∗
(0.009)
−0.079
(0.091)
−0.293
(0.493)
3,611.065
0.003
(0.563)
0.579∗∗∗
(0.205)
1.064∗∗∗
(0.114)
−0.320∗∗
(0.151)
0.040
(0.063)
0.240
(0.198)
0.073
(0.149)
−0.004
(0.011)
−1.918∗∗∗
(0.391)
−5.689∗∗∗
(0.999)
3,611.065
−0.134
(0.429)
0.536∗∗∗
(0.160)
0.512∗∗∗
(0.075)
−0.364∗∗∗
(0.125)
−0.090∗
(0.046)
0.097
(0.158)
0.320∗∗∗
(0.116)
0.010
(0.008)
−1.296∗∗∗
(0.129)
−1.597∗∗∗
(0.523)
3,611.065
2.139∗∗∗
(0.605)
0.806∗∗∗
(0.205)
0.443∗∗∗
(0.103)
0.211
(0.195)
−0.161∗∗
(0.063)
−0.460∗
(0.235)
0.366∗∗
(0.157)
0.033∗∗∗
(0.012)
0.103
(0.134)
−1.432∗∗
(0.729)
3,611.065
0.651
(0.468)
0.236
(0.175)
−0.391∗∗∗
(0.087)
−0.093
(0.155)
−0.141∗∗∗
(0.051)
0.272
(0.172)
0.436∗∗∗
(0.133)
0.048∗∗∗
(0.010)
0.300∗∗
(0.118)
−2.322∗∗∗
(0.559)
3,611.065
0.567
(0.430)
0.411∗∗
(0.161)
0.020
(0.075)
0.100
(0.132)
−0.163∗∗∗
(0.047)
0.109
(0.155)
−0.172
(0.117)
0.017∗
(0.009)
−0.197∗∗
(0.092)
−1.365∗∗∗
(0.497)
3,611.065
−0.046
(0.302)
0.089
(0.118)
0.239∗∗∗
(0.054)
−0.135
(0.095)
−0.062∗
(0.034)
0.219∗
(0.113)
0.051
(0.084)
0.016∗∗
(0.006)
−0.038
(0.066)
0.392
(0.355)
3,611.065
p < .05;
∗∗∗
p < .01
93
8.5
Appendix E: MCMC Diagnostics
Sensitivity Analysis
The prior for the regression coefficients is a multivariate normal with p parameters:
β ∼ Np (~µ = ~0, Λ = 10−1 ∗ Ipxp ), where Λ = Σ−1 is the precision matrix. The density is
as follows:
1
1
~ )T Λ(X − µ
~ )}
f (X | µ
~ , Λ) =
exp{− (X − µ
np/2
n/2
(2π)
|Λ|
2
As the sample size increases, the influence of the prior on the posterior decreases; Bayesian
asymptotics concords with frequentist, following the Law of Large Numbers and Central
Limit Theorem. As an example, for a normal distribution in which σ is known,
β ∼ N (µ, σ). The prior is β0 ∼ N (µ0 , σ0 ), leading to an updated posterior of
βp ∼ N (µp , σp ), where:
N
σ2
N σ02
1 X
µp =
µo +
∗
xi
N σ02 + σ 2
N σ02 + σ 2 N i=1
σp2 = (
1
N
+ 2 )−1
2
σ0 σ
(5)
(6)
From (7), as n → ∞, µp → µM LE , where µM LE is the Maximum Likelihood Estimate,
PN
1
2
i=1 xi . Following from (8), as n → ∞, σp → 0. Similar to the MLE, the Bayesian
N
estimators are asymptotically unbiased and efficient, converging to a normal distribution.
Given that N > 6, 000 for my analysis, µp ≈ µM LE .
94
Diagnostics Results
Vote Switch Diagnostics
Intercept
Efficacy
Satisfaction
Policy
Income
Class
Education
Age
Geweke
.2, .2
-0.624
-0.457
-1.144
0.824
0.675
-0.471
-0.137
0.485
Statistics
.4, .4
.1, .5
-0.792 -0.606
-1.614 -0.689
-0.031 -0.050
0.859 -0.565
0.764
1.163
-0.748 -0.599
0.767
0.490
-0.374 -0.069
.5, .1
-2.343
-0.358
-0.548
1.308
0.371
0.578
0.540
0.894
Heidelberger–Welch Test
Stationarity CVM
Intercept
0.360
0.001
Efficacy
0.199
0.001
Satisfaction
0.790
0.001
Policy
0.310
0.001
Income
0.623
0.001
Class
0.450
0.002
Education
0.471
0.001
Age
0.949
0.001
Cross-Correlation
Intercept
Efficacy
Satisfaction
Policy
Income
Class
Education
Age
Lag
Lag
Lag
Lag
Lag
0
20
100
200
1000
Intercept
1.000
-0.040
0.096
-0.017
-0.037
-0.041
-0.277
-0.092
Intercept
1.000
0.243
-0.013
0.007
0.008
Efficacy
-0.040
1.000
-0.157
0.110
-0.031
-0.100
-0.100
0.196
Efficacy
1.000
0.246
0.016
0.001
0.000
Satisfaction
0.096
-0.157
1.000
-0.211
-0.037
-0.085
0.035
-0.011
Policy
-0.017
0.110
-0.211
1.000
-0.031
-0.052
0.043
-0.138
Autocorrelation
Satisfaction
Policy
1.000
1.000
0.251
0.244
-0.005
0.012
-0.013
0.002
0.002
0.001
95
Income
-0.037
-0.031
-0.037
-0.031
1.000
-0.264
-0.164
0.096
Income
1.000
0.242
0.003
0.000
-0.011
Class
-0.041
-0.100
-0.085
-0.052
-0.264
1.000
-0.406
-0.122
Class
1.000
0.241
0.004
-0.018
-0.002
Education
-0.277
-0.100
0.035
0.043
-0.164
-0.406
1.000
0.153
Education
1.000
0.233
0.010
0.005
0.020
Age
-0.092
0.196
-0.011
-0.138
0.096
-0.122
0.153
1.000
Age
1.000
0.253
-0.012
-0.019
-0.013
Mainstream Vote Diagnostics
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
Geweke
.2, .2
-0.952
1.445
-0.363
0.809
0.701
-0.603
0.000
0.894
Statistics
.4, .4
.1, .5
-0.794 -0.954
-0.039 0.695
-0.031 0.030
0.450 -0.135
0.512
0.589
-0.169 0.561
0.383
0.178
-0.283 1.006
.5, .1
-1.718
2.079
1.476
0.105
-0.385
0.951
-0.412
1.253
Only variables with a statistics greater than two included
Heidelberger–Welch
Stationarity CVM
Intercept
0.455
0.001
Efficacy
0.248
0.001
Satisfaction
0.750
0.001
Left–Right
0.897
0.001
Income
0.879
0.001
Class
0.685
0.002
Education
0.878
0.001
Age
0.543
0.001
Cross–Correlation
Intercept
Efficacy
Satisfaction
Left–Right
Income
Class
Education
Age
Lag
Lag
Lag
Lag
Lag
0
20
100
200
1000
Intercept
1.000
-0.089
0.116
0.128
-0.012
-0.044
-0.292
-0.103
Intercept
1.000
0.255
-0.004
0.003
-0.016
Efficacy
-0.089
1.000
-0.131
0.091
-0.034
-0.102
-0.081
0.226
Efficacy
1.000
0.246
-0.007
0.004
0.006
Satisfaction
0.116
-0.131
1.000
-0.216
-0.016
-0.057
-0.032
-0.002
Left–Right
0.128
0.091
-0.216
1.000
-0.054
-0.141
0.129
0.009
Autocorrelation
Satisfaction
Left–Right
1.000
1.000
0.250
0.240
0.005
-0.000
-0.007
-0.009
0.002
0.002
96
Income
-0.012
-0.034
-0.016
-0.054
1.000
-0.248
-0.186
0.072
Income
1.000
0.238
-0.007
-0.001
-0.017
Class
-0.044
-0.102
-0.057
-0.141
-0.248
1.000
-0.412
-0.100
Class
1.000
0.233
0.005
0.011
-0.005
Education
-0.292
-0.081
-0.032
0.129
-0.186
-0.412
1.000
0.139
Education
1.000
0.240
0.008
0.002
0.002
Age
-0.103
0.226
-0.002
0.009
0.072
-0.100
0.139
1.000
Age
1.000
0.256
-0.003
0.007
-0.000
Mainstream Party Choice Diagnostics
(Intercept).CDA
(Intercept).PvdA
(Intercept).VVD
Efficacy.CDA
Efficacy.PvdA
Efficacy.VVD
Satisfaction.CDA
Satisfaction.PvdA
Satisfaction.VVD
Left–Right.CDA
Left–Right.PvdA
Left–Right.VVD
Policy.CDA
Policy.PvdA
Policy.VVD
Income.CDA
Income.PvdA
Income.VVD
Class.CDA
Class.PvdA
Class.VVD
Education.CDA
Education.PvdA
Education.VVD
Age.CDA
Age.PvdA
Age.VVD
Geweke Statistic
g1.z
g2.z
-0.926
-1.010
-0.601
-0.928
-0.799
-0.448
1.151
0.676
1.309
1.221
1.375
2.024
0.567
1.257
-0.431
-0.396
-0.349
0.404
0.358
0.155
1.186
-0.212
0.509
-0.411
0.487
1.643
0.546
1.894
0.240
-0.065
-0.848
-1.620
-0.679
0.245
-0.707
-0.948
0.115
-0.162
1.024
0.779
1.281
0.818
-1.116
-0.909
0.847
-0.041
-1.051
-1.244
0.558
0.291
1.230
0.860
0.285
1.141
g3.z
-0.569
0.200
0.264
1.135
0.278
0.624
0.045
-0.162
0.172
0.144
1.032
-0.652
1.062
0.458
1.243
-1.339
-1.366
-1.034
-1.540
-1.171
-0.196
0.212
1.575
0.168
1.265
0.914
0.308
Heidelberger–Welch
Stationarity
(Intercept).CDA
0.430
(Intercept).PvdA
0.294
(Intercept).VVD
0.527
Efficacy.CDA
0.526
Efficacy.PvdA
0.474
Efficacy.VVD
0.360
Satisfaction.CDA
0.415
Satisfaction.PvdA
0.708
Satisfaction.VVD
0.985
Left–Right.CDA
0.319
Left–Right.PvdA
0.387
Left–Right.VVD
0.976
Policy.CDA
0.676
Policy.PvdA
0.620
Policy.VVD
0.434
Income.CDA
0.694
Income.PvdA
0.604
Income.VVD
0.609
Class.CDA
0.836
Class.PvdA
0.458
Class.VVD
0.857
Education.CDA
0.486
Education.PvdA
0.770
Education.VVD
0.148
Age.CDA
0.735
Age.PvdA
0.636
Age.VVD
0.782
g4.z
-0.705
0.927
0.958
0.554
1.067
0.927
0.909
0.178
0.046
1.244
1.199
0.364
0.113
0.435
0.211
-0.252
0.222
-0.903
-0.345
0.791
0.285
-1.013
-0.314
-1.926
0.621
0.223
0.499
CVM
0.001
0.001
0.001
0.001
0.001
0.002
0.001
0.001
0.002
0.002
0.002
0.002
0.002
0.002
0.002
0.001
0.001
0.002
0.002
0.002
0.002
0.002
0.002
0.002
0.001
0.001
0.002
Cross–Correlation
Intercept
Efficacy
Satisfaction
Left–Right
Policy
Income
Class
Education
Age
Intercept
1.000
-0.084
-0.127
-0.252
-0.122
-0.062
-0.027
-0.204
-0.030
Efficacy
-0.084
1.000
-0.160
0.010
0.087
-0.044
-0.109
-0.078
0.212
Satisfaction
-0.127
-0.160
1.000
-0.052
-0.045
-0.015
-0.056
0.010
0.012
Left–Right
-0.252
0.010
-0.052
1.000
-0.163
0.030
-0.010
0.102
0.005
Policy
-0.122
0.087
-0.045
-0.163
1.000
0.015
-0.001
-0.025
-0.064
Income
-0.062
-0.044
-0.015
0.030
0.015
1.000
-0.204
-0.200
0.056
Class
-0.027
-0.109
-0.056
-0.010
-0.001
-0.204
1.000
-0.378
-0.084
Education
-0.204
-0.078
0.010
0.102
-0.025
-0.200
-0.378
1.000
0.094
Age
-0.030
0.212
0.012
0.005
-0.064
0.056
-0.084
0.094
1.000
Only CDA cross–correlations
Autocorrelation
Lag
Lag
Lag
Lag
Lag
0
20
100
200
1000
Intercept
1.000
0.015
0.013
0.002
-0.006
Efficacy
1.000
0.009
0.005
0.006
0.002
Satisfaction
1.000
-0.010
-0.010
-0.018
-0.004
Left–Right
1.000
0.002
0.001
-0.007
0.010
Policy
1.000
0.000
0.002
-0.003
-0.007
Only CDA autocorrelations
97
Income
1.000
0.005
0.014
-0.005
-0.000
Class
1.000
0.001
0.009
-0.003
0.005
Education
1.000
-0.020
0.001
-0.010
0.000
Age
1.000
0.006
-0.001
0.015
0.006
Outsider Party Choice Diagnostics
Geweke Statistics
.2, .2
.4, .4
2.125
1.704
-1.212
-0.262
1.660
0.282
2.305
1.709
2.038
-0.631
-1.185
-1.187
1.221
-0.016
-1.500
-1.289
-2.376
-0.850
-2.059
-0.803
-1.706
-0.330
Efficacy.Other
Satisfaction.GroenLinks
Policy.D66
Policy.GroenLinks
Policy.Other
Class.Other
Education.ChristenUnie
Education.GroenLinks
Education.SP
Age.D66
Age.LPF
.1, .5
0.558
-0.873
1.903
1.336
0.568
-0.311
2.013
-2.532
-1.318
-2.211
-3.183
.5, .1
0.864
-2.404
2.412
1.569
1.140
-2.103
1.236
-2.479
-1.350
-2.742
-1.119
Only variables with a statistics greater than two included
Heidelberger–Welch
Stationarity CVM
Class.ChristenUnie
0.350
0.003
Age.Other
0.962
0.002
All variables passed stationarity test;
Only variables that failed half–width test included.
Intercept
Efficacy
Satisfaction
Left–Right
Policy
Income
Class
Education
Age
Lag
Lag
Lag
Lag
Lag
0
20
100
200
1000
Intercept
1.000
-0.185
0.147
-0.371
-0.463
0.039
-0.006
-0.369
0.017
Intercept
1.000
0.020
-0.010
0.006
-0.000
Efficacy
-0.185
1.000
-0.195
0.055
0.080
-0.016
-0.074
-0.060
0.194
Efficacy
1.000
0.022
-0.008
0.013
0.009
Cross–Correlation
Satisfaction Left–Right Policy Income
0.147
-0.371
-0.463
0.039
-0.195
0.055
0.080
-0.016
1.000
-0.143
0.009
0.031
-0.143
1.000
-0.296
0.025
0.009
-0.296
1.000
-0.040
0.031
0.025
-0.040
1.000
-0.061
-0.062
-0.043
-0.214
0.005
0.181
0.098
-0.197
0.025
0.010
-0.095
0.073
Only Christen Unie cross–correlations
Autocorrelation
Satisfaction Left–Right Policy Income
1.000
1.000
1.000
1.000
0.016
0.012
0.018
0.024
0.016
-0.006
0.001
-0.011
-0.000
0.000
-0.000
0.001
-0.007
0.008
0.018
0.006
Only Christen Unie autocorrelations
98
Class
-0.006
-0.074
-0.061
-0.062
-0.043
-0.214
1.000
-0.367
-0.092
Class
1.000
0.033
-0.003
-0.003
-0.012
Education
-0.369
-0.060
0.005
0.181
0.098
-0.197
-0.367
1.000
0.150
Education
1.000
0.028
-0.010
0.009
-0.007
Age
0.017
0.194
0.025
0.010
-0.095
0.073
-0.092
0.150
1.000
Age
1.000
0.030
0.006
0.003
-0.009
8.6
Appendix F: Diagnostic Test Statistics
Geweke Statistic: Uses two specified portions of the Markov Chain—one at the beginning and the other from
the end—and compares the two sections of the chain’s mean to test for convergence. If the two means
are similar, then the chain has converged. I ran the Geweke diagnostic four times per model, with the
beginning and ending quantiles specified as follows: .2, .2; .4, .4; .1, .5; and .5, .1.
Let {θt } be a Markov Chain, with {θ1t : t = 1, . . . , n1 } and {θ2t : t = na , . . . , n} two portions of the chain,
where 1 < n1 < na < n, and set n2 = n − nP
a + 1. Let ŝ1 (0) and ŝ1 (0) be the respective spectral densities
Pn
t
t
1
1
¯
estimated at zero frequency. Then θ¯1 = n11 n
i=1 θ and θ2 = n2
i=na θ .
The Geweke Statistic is the difference of means test:
Zn = q
θ¯1 − θ¯2
ŝ1 (0)
n1
+
ŝ2 (0)
n2
The null hypothesis—equality of distributions—is rejected with a large absolute value of the z-score, that
is, a z-score greater than 1.96. (Cowles & Carlin 1996, Geweke 1991).
Heidelberger–Welch Diagnostic: A two–part test. First, it test whether the Markov Chain is from a stationary distribution, or, for a time series, X(t): 1) E[X(t)2 ] < ∞, 2) E[X(t)] = µ ∀t, and 3) γ(t, t + h) = γ(h).
Second, it tests whether the chain has adequate iterations to estimate the posterior mean. The test uses
a (1 − α) confidence interval about the mean, dividing the half–width by the sample mean, based on the
Cramer–von Mises test statistic:
Z ∞
ωn2 =
(Fn (x) − F (x))2 dF (x)
−∞
t
Given {θ }, set S0 = 0 and Sn =
Pn
t=1
t
θ . Construct a sequence:
Bn (S) =
S[ns] − [ns]θ̄
p
np̂()
With [] the rounding
operator and p̂() the spectral density estimated at zero frequency.
R1
As n → ∞, 0 Bn (S)2 ds = CV M (Bn ), the standard Cramer–von Mises distribution.
The iterations, less the stationary portion of the chain—that which passed the first test—is tested with
the half–width test. For a (1 − α) significance level, the test statistics is:
q
Z1− α2 ( snˆn )
θ̂
A variables passes the second test if the proportion between the two is less than some (Cowles &
Carlin 1996, Heidelberger & Welch 1981, Heidelberger & Welch 1983).
Autocorrelation: Given a time series θt , the autocorrelation function (ACF) between θt and θt+h is given by:
ρ̂(h) =
γ̂(h)
γ̂(0)
Where the auto–covariance function (ACVF) is the covariance between the current iteration and an iterations h steps ahead. The ACVF is estimated by:
γ̂(h) =
n−|h|
1 X t+h
(θ
− θ̄t )(θt − θ̄t )
n t=1
Effective Sample Size: Tests how well the Markov Chain mixes. Given the autocorrelation, the ESS is the
effective amount of information gained from the iterations. A high ACF implies that each iteration
contributes less information.
n
P
ESS =
1+2 ∞
k=1 ρk (θ)
99
References
Achen, C. H. (2006). Expressive bayesian voters, their turnout decisions, and double probit:
Empirical implications of a theoretical model, Princeton University manuscript. Downloaded from the Society for Political Methodology: http://polmeth. wustl. edu/retrieve. php
.
Arzheimer, K. (2009). Contextual factors and the extreme right vote in western europe, 1980–
2002, American Journal of Political Science 53(2): 259–275.
Arzheimer, K. & Carter, E. (2006). Political opportunity structures and right-wing extremist
party success, European Journal of Political Research 45(3): 419–443.
Bale, T. (2003). Cinderella and her ugly sisters: the mainstream and extreme right in europe’s
bipolarising party systems, West European Politics 26(3): 67–90.
Barr, R. R. (2009). Populists, outsiders and anti-establishment politics, Party Politics 15(1): 29–
48.
Bartels, B. L. (2009). The constraining capacity of legal doctrine on the us supreme court,
American Political Science Review 103(03): 474–495.
Bayarri, M. J. & Berger, J. O. (2004). The interplay of bayesian and frequentist analysis,
Statistical Science pp. 58–80.
Bayes, T. & Price, R. (1763). An essay towards solving a problem in the doctrine of chances. by
the late rev. mr. bayes, frs communicated by mr. price, in a letter to john canton, amfrs,
Philosophical Transactions (1683-1775) pp. 370–418.
Bendor, J., Mookherjee, D., Ray, D. et al. (2006). Satisficing and selection in electoral competition, Quarterly Journal of Political Science 1(2): 171–200.
Berger, J. O. (1982). The robust Bayesian viewpoint, Purdue University. Department of Statistics.
Berger, J. O. & Molina, G. (2005). Posterior model probabilities via path-based pairwise priors,
Statistica Neerlandica 59(1): 3–15.
Bernardo, J. M. (1984). Monitoring the 1982 spanish socialist victory: a bayesian analysis,
Journal of the American Statistical Association 79(387): 510–515.
Bonneau, C. W. & Cann, D. M. (2011). Campaign spending, diminishing marginal returns, and
campaign finance restrictions in judicial elections, The Journal of Politics 73(04): 1267–
1280.
Campbell, A., Gurin, G. & Miller, W. E. (1954). The voter decides., Row, Peterson, and Co.
Capoccia, G. (2002). Anti-system parties a conceptual reassessment, Journal of Theoretical
Politics 14(1): 9–35.
Carmines, E. G. & Stimson, J. A. (1980). The two faces of issue voting, American Political
Science Review 74(01): 78–91.
Clarke, H. D. & Acock, A. C. (1989). National elections and political attitudes: The case of
political efficacy, British Journal of Political Science 19(04): 551–562.
Clyde, M. A., Ghosh, J. & Littman, M. L. (2012). Bayesian adaptive sampling for variable
selection and model averaging, Journal of Computational and Graphical Statistics .
100
Coffé, H. (2005). Do individual factors explain the different success of the two belgian extreme
right parties, Acta Politica 40(1): 74–93.
Coleman, K. M. & Davis, C. L. (1976). The structural context of politics and dimensions of
regime performance:” their importance for the comparative study of political efficacy”,
Comparative Political Studies 9(2): 189.
Conradt, D. P. & Dalton, R. J. (1988). The west german electorate and the party system:
Continuity and change in the 1980s, The Review of Politics 50(01): 3–29.
Cowles, M. K. & Carlin, B. P. (1996). Markov chain monte carlo convergence diagnostics: a
comparative review, Journal of the American Statistical Association 91(434): 883–904.
Craig, S. C. (1980). The mobilization of political discontent, Political Behavior 2(2): 189–209.
Craig, S. C. & Maggiotto, M. A. (1981). Political discontent and political action, The Journal
of Politics 43(02): 514–522.
Dalton, R. J. (2000). Parties without Partisans. Political Change in Advanced Industrial Democracies. Oxford University Press, Oxford, Oxford University Press, chapter Unthinkable
Democracy: Political Change in Advanced Industrial Democracies.
de Moor, J. (2014). Reinterpreting the relation between external efficacy and political participation, Paper presented at the 5th ECPR General Conference, Vol. 4, p. 6.
Downs, A. (1957). An economic theory of political action in a democracy, The journal of political
economy pp. 135–150.
Erlingsson, G. Ó. & Persson, M. (2011). The swedish pirate party and the 2009 european
parliament election: protest or issue voting?, Politics 31(3): 121–128.
Fennema, M. (1997). Some conceptual issues and problems in the comparison of anti-immigrant
parties in western europe, Party Politics 3(4): 473–492.
Ferejohn, J. A. & Fiorina, M. P. (1974). The paradox of not voting: A decision theoretic
analysis, American political science review 68(02): 525–536.
Finkel, S. E. (1985). Reciprocal effects of participation and political efficacy: A panel analysis,
American Journal of political science pp. 891–913.
Finkel, S. E. & Muller, E. N. (1998). Rational choice and the dynamics of collective political
action: Evaluating alternative models with panel data, American Political Science Review
92(01): 37–49.
Franklin, M., Niemi, R. & Whitten, G. (1994). The two faces of tactical voting, British Journal
of Political Science 24(04): 549–557.
Gamson, W. A. (1968). Power and Discontent, The Dorsey Press.
Gelman, A., Carlin, J. B., Stern, H. S. & Rubin, D. B. (2014). Bayesian data analysis, Vol. 2,
Taylor & Francis.
George, E. I. (2000). The variable selection problem, Journal of the American Statistical Association 95(452): 1304–1308.
Gerber, A. S., Green, D. P. & Shachar, R. (2003). Voting may be habit-forming: evidence from
a randomized field experiment, American Journal of Political Science 47(3): 540–550.
101
Geweke, J. (1991). Evaluating the accuracy of sampling-based approaches to the calculation of
posterior moments, Technical report, Federal Reserve Bank of Minneapolis.
Gill, J. (2014). Bayesian Methods: A Social and Behavioral Sciences Approach, hapman and
Hall.
Gill, J. & Walker, L. D. (2005). Elicited priors for bayesian model specifications in political
science research, Journal of Politics 67(3): 841–872.
Golder, M. (2016). Extreme right parties in europe, Annual Review of Political Science 19(1).
Grönlund, K. & Setälä, M. (2007). Political trust, satisfaction and voter turnout, Comparative
European Politics 5(4): 400–422.
Harsanyi, J. C. (1969). Rational-choice models of political behavior vs. functionalist and conformist theories, World politics 21(04): 513–538.
Hastings, W. K. (1970). Monte carlo sampling methods using markov chains and their applications, Biometrika 57(1): 97–109.
Heidelberger, P. & Welch, P. D. (1981). A spectral method for confidence interval generation
and run length control in simulations, Communications of the ACM 24(4): 233–245.
Heidelberger, P. & Welch, P. D. (1983). Simulation run length control in the presence of an
initial transient, Operations Research 31(6): 1109–1144.
Hix, S. & Marsh, M. (2007). Punishment or protest? understanding european parliament
elections, Journal of Politics 69(2): 495–510.
Hoeting, J. A., Madigan, D., Raftery, A. E. & Volinsky, C. T. (1999). Bayesian model averaging:
a tutorial, Statistical science pp. 382–401.
Hooghe, M. & Dassonneville, R. (2014). The spiral of distrust. a panel study on the relation
between political distrust and protest voting in belgium.
Hooghe, M. & Marien, S. (2013). A comparative analysis of the relation between political trust
and forms of political participation in europe, European Societies 15(1): 131–152.
Hooghe, M., Marien, S. & Pauwels, T. (2011). Where do distrusting voters turn if there is
no viable exit or voice option? the impact of political trust on electoral behaviour in the
belgian regional elections of june 2009 1, Government and Opposition 46(2): 245–273.
Ignazi, P. (1996). The crisis of parties and the rise of new political parties, Party Politics
2(4): 549–566.
Inglehart, R. (1981). Post-materialism in an environment of insecurity, American Political
Science Review 75(04): 880–900.
Inglehart, R. & Flanagan, S. C. (1987). Value change in industrial societies, American Political
Science Review 81(04): 1289–1319.
Jessee, S. A. (2012). Ideology and spatial voting in American elections, Cambridge University
Press.
Katz, R. S. & Mair, P. (1995). Changing models of party organization and party democracy
the emergence of the cartel party, Party politics 1(1): 5–28.
Katz, R. S. & Mair, P. (2009). The cartel party thesis: A restatement, Perspectives on politics
7(04): 753–766.
102
Keman, H. (2014). Democratic performance of parties and legitimacy in europe, West European
Politics 37(2): 309–330.
Klandermans, B. (1996). The social psychology of protest.
Kollman, K., Miller, J. H. & Page, S. E. (1992). Adaptive parties in spatial elections, American
Political Science Review 86(04): 929–937.
Kselman, D. & Niou, E. (2011). Protest voting in plurality elections: a theory of voter signaling,
Public Choice 148(3-4): 395–418.
Labzina, E. & Schofield, N. (2015). Valence and ideological proximity in the rise of nationalism:
Party support in the spanish general elections, 2008 and 2011.
Laplace, P.-S. (2012). Pierre-Simon Laplace Philosophical Essay on Probabilities: Translated
from the fifth French edition of 1825 With Notes by the Translator, Vol. 13, Springer Science
& Business Media.
Lecoutre, B., Lecoutre, M.-P. & Poitevineau, J. (2001). Uses, abuses and misuses of significance
tests in the scientific community: Won’t the bayesian choice be unavoidable?, International
Statistical Review 69(3): 399–417.
Levin, M. B. & Eden, M. (1962). Political strategy for the alienated voter, Public Opinion
Quarterly 26(1): 47–63.
Lindley, D. V. & Smith, A. F. (1972). Bayes estimates for the linear model, Journal of the
Royal Statistical Society. Series B (Methodological) pp. 1–41.
McAdam, D. & Tarrow, S. (2010). Ballots and barricades: On the reciprocal relationship
between elections and social movements, Perspectives on Politics 8(02): 529–542.
Metropolis, N., Rosenbluth, A. W., Rosenbluth, M. N., Teller, A. H. & Teller, E. (1953).
Equation of state calculations by fast computing machines, The journal of chemical physics
21(6): 1087–1092.
Metropolis, N. & Ulam, S. (1949). The monte carlo method, Journal of the American statistical
association 44(247): 335–341.
Michels, R. (1915). Political parties: A sociological study of the oligarchical tendencies of modern
democracy, Hearst’s International Library Company.
Middendorp, C. P. (1992). Left-right self-identification and (post) materialism in the ideological
space; their effect on the vote in the netherlands, Electoral Studies 11(3): 249–260.
Miller, A. H. (1974). Political issues and trust in government: 1964–1970., American Political
Science Review 68(03): 951–972.
Mokken, R. J. (1971). A theory and procedure of scale analysis: With applications in political
research, Vol. 1, Walter de Gruyter.
Moller, T. (1999). The swedish election 1998: A protest vote and the birth of a new political
landscape?, Scandinavian Political Studies 22(3): 261–276.
Mudge, S. L. & Chen, A. S. (2014). Political parties and the sociological imagination: Past,
present, and future directions, Annual Review of Sociology 40: 305–330.
Muller, E. N. & Jukam, T. O. (1977). On the meaning of political support, American Political
Science Review 71(04): 1561–1595.
103
Myatt, D. P. (2015). A theory of protest voting, The Economic Journal .
Nachmias, D. (1974). Modes and types of political alienation, The British Journal of Sociology
25(4): 478–493.
Niemi, R. G., Craig, S. C. & Mattei, F. (1991). Measuring internal political efficacy in the 1988
national election study., American Political Science Review 85(04): 1407–1413.
Nurmi, H. (1999). Voting paradoxes and how to deal with them, Springer Science & Business
Media.
Oberschall, A. (1978). Theories of social conflict, Annual review of sociology pp. 291–315.
Ordeshook, P. C. (1976). Party Identification and Beyond: Representations of Voting and Party
Competition, ECPR Press, chapter The spatial theory of elections: A review and a critique.
Pahre, R. & Radziszewski, E. (2006). Euroskeptic voting in european and national elections,
Urbana 51: 61801.
Paige, J. M. (1971). Political orientation and riot participation, American Sociological Review
pp. 810–820.
Pattie, C. & Johnston, R. (1998). Voter turnout at the british general election of 1992: Rational choice, social standing or political efficacy?, European Journal of Political Research
33(2): 263–283.
Pollock III, P. H. (1983). The participatory consequences of internal and external political
efficacy: A research note, The Western Political Quarterly pp. 400–409.
Quattrone, G. A. & Tversky, A. (1988). Contrasting rational and psychological analyses of
political choice, American Political Science Review 82(03): 719–736.
Quinn, K. M. & Martin, A. D. (2002). An integrated computational model of multiparty
electoral competition, Statistical Science pp. 405–419.
Quinn, K. M., Martin, A. D. & Whitford, A. B. (1999). Voter choice in multi-party democracies:
a test of competing theories and models, American Journal of Political Science pp. 1231–
1247.
Richey, M. (2010). The evolution of markov chain monte carlo methods, The American Mathematical Monthly 117(5): 383–413.
Riker, W. H. & Ordeshook, P. C. (1968). A theory of the calculus of voting, American political
science review 62(01): 25–42.
Rosas, G. (2006). Bagehot or bailout? an analysis of government responses to banking crises,
American Journal of Political Science 50(1): 175–191.
Scott, J. G., Berger, J. O. et al. (2010). Bayes and empirical-bayes multiplicity adjustment in
the variable-selection problem, The Annals of Statistics 38(5): 2587–2619.
Seeman, M. (1959). On the meaning of alienation, American sociological review pp. 783–791.
Seligson, M. A. (1980). Trust, efficacy and modes of political participation: A study of costa
rican peasants, British Journal of Political Science 10(01): 75–98.
Shotts, K. W. (2006). A signaling model of repeated elections, Social choice and Welfare
27(2): 251–261.
104
Sigelman, L. & Feldman, S. (1983). Efficacy, mistrust, and political mobilization a cross-national
analysis, Comparative Political Studies 16(1): 118–143.
Spoon, J.-J. & Hobolt, S. B. (2010). Motivating the european voter: Parties, issues, and
campaigns, APSA 2010 Annual Meeting Paper.
Superti, C. (2014). Filling in the blanks: Understanding blank and null voting as a form
of informed protest and a new metric of public discontent, Midwestern Political Science
Association.
Thijssen, P. & Dierickx, G. (2001). The extreme right and political alienation: A causality riddle.
the case of the vlaams blok in belgium, ECPR Joint Sessions of Workshops, Grenoble.
Thompson, W. E. & Horton, J. E. (1960). Political alienation as a force in political action,
Social Forces pp. 190–195.
Treib, O. (2014). The voter says no, but nobody listens: causes and consequences of
the eurosceptic vote in the 2014 european elections, Journal of European Public Policy
21(10): 1541–1554.
Tsangarides, M. C. G. (2004). A Bayesian approach to model uncertainty, number 4-68, International Monetary Fund.
Valentino, N. A., Gregorowicz, K. & Groenendyk, E. W. (2009). Efficacy, emotions and the
habit of participation, Political Behavior 31(3): 307–330.
Van der Brug, W. & Fennema, M. (2003). Protest or mainstream? how the european antiimmigrant parties have developed into two separate groups by 1999, European Journal of
Political Research 42(1): 55–76.
Van der Brug, W. & Fennema, M. (2007). Causes of voting for the radical right, International
Journal of Public Opinion Research 19(4): 474–487.
Van der Brug, W., Fennema, M. & Tillie, J. (2000). Anti-immigrant parties in europe: Ideological or protest vote?, European Journal of Political Research 37(1): 77–102.
Van der Brug, W., Fennema, M. & Tillie, J. (2005). Why some anti-immigrant parties fail
and others succeed a two-step model of aggregate electoral support, Comparative Political
Studies 38(5): 537–573.
Van der Eijk, C., Franklin, M. & Marsh, M. (1996). What voters teach us about europe-wide
elections: what europe-wide elections teach us about voters, Electoral Studies 15(2): 149–
166.
Van Der Eijk, C., Irwin, G. & Niemöller, K. (1986). The dutch parliamentary election of may
1986, Electoral Studies 5(3): 289–296.
Van Holsteyn, J. J. & Irwin, G. A. (2003). Never a dull moment: Pim fortuyn and the dutch
parliamentary election of 2002, West European Politics 26(2): 41–66.
Vasilopoulos, P. & Demertzis, N. (2013). The greek green voter: environmentalism or protest?,
Environmental Politics 22(5): 728–738.
Wasserman, L. (2000). Bayesian model selection and model averaging, Journal of mathematical
psychology 44(1): 92–107.
Weihs, C., Mersmann, O. & Ligges, U. (2013). Foundations of Statistical Algorithms: With
References to R Packages, CRC Press.
105
Whiteley, P. F. (1995). Rational choice and political participationevaluating the debate, Political
Research Quarterly 48(1): 211–233.
Wolfsfeld, G. (1985). Political efficacy and political action: A change in focus using data from
israel, Social Science Quarterly 66(3): 617.
Wolfsfeld, G. (1986). Political action repertoires the role of efficacy, Comparative Political
Studies 19(1): 104–129.
Zaslove, A. (2008). Here to stay? populism as a new party type, European review 16(03): 319–
336.
Zimmerman, M. A., Israel, B. A., Schulz, A. & Checkoway, B. (1992). Further explorations
in empowerment theory: An empirical analysis of psychological empowerment, American
Journal of Community Psychology 20(6): 707–727.
106